Linked to DIYguru's ADAS & AUTOSAR nanodegree
100 ADAS & AUTOSAR Projects for Final Year, M.Tech & PhD
From sensor perception to ECU integration: camera, radar and lidar perception, sensor fusion and tracking, adaptive cruise and emergency braking, lane keeping, parking and path planning, scenario-based testing, AUTOSAR Classic software components, AUTOSAR Adaptive and service-oriented architecture, functional safety and cybersecurity. Each guide lists the tools and datasets you need, the steps with checkpoints, how to test against scenarios and metrics, and what to report.
40
B.Tech / BE projects
33
M.Tech projects
27
PhD projects
Need help choosing or building your project?
Tell us your branch, level and the topic you like. Our team replies with mentoring, course and workshop options.
Final-year B.Tech / BE: buildable in one semester by a team of 3–4, with a working demo and clear results.
M.Tech: modelling, a baseline to compare against and quantified results for your thesis.
PhD: a method validated on real models or hardware, with research questions, baselines and metrics.
Project ideas
Pick a project, then follow the guide
Filter by your level and the area you want to work in. Open any project for its full step-by-step guide.
- B.Tech / BECamera perceptionBeginner
1.Lane Detection with OpenCV: Canny, Hough Transform and Lane Fitting
Build a classical lane-detection pipeline in Python and OpenCV that finds lane markings with colour thresholds, Canny edges and the probabilistic Hough transform, then fits and tracks smooth lane lines across video frames. It is scored against labelled frames from Indian and public road footage.
4 weeksPython · OpenCV - B.Tech / BECamera perceptionIntermediate
2.Traffic Sign Recognition with a CNN on Indian and GTSRB Signs
Train a convolutional neural network that classifies traffic signs from GTSRB and a small self-collected set of Indian signs, then measure how well it transfers between the two. The project covers data handling, training, augmentation, confusion analysis and a simple detect-then-classify demo.
6 weeksPython · PyTorch or TensorFlow - B.Tech / BECamera perceptionIntermediate
3.Pedestrian and Vehicle Detection with YOLO on Indian Road Video
Fine-tune a YOLO detector to find pedestrians, two-wheelers, auto-rickshaws, cars and trucks in Indian road video using IDD, then evaluate it with mAP, per-class recall and frame rate. The project shows how a general detector is adapted to local road users and how its errors matter for ADAS.
6 weeksPython · YOLO · OpenCV - B.Tech / BECamera perceptionIntermediate
4.Driver Drowsiness and Distraction Detection from a Cabin Camera
Build a cabin-camera driver monitor in Python that tracks face landmarks with MediaPipe, computes eye closure (PERCLOS), blink and yawn measures and head pose, and raises staged drowsiness and distraction alerts. It is tested on recorded sessions with labelled events and measured for false alarms.
5 weeksPython · OpenCV · MediaPipe - B.Tech / BECamera perceptionIntermediate
5.Monocular Camera Distance Estimation to the Vehicle Ahead
Estimate the distance to the vehicle ahead from a single camera using a calibrated pinhole model, the flat-road ground-contact method and a known-width method, then compare their errors against ground truth from KITTI. The project also derives time to collision from the distance track.
5 weeksPython · OpenCV - M.TechCamera perceptionAdvanced
6.Drivable-Area Segmentation for Unstructured Indian Roads
Train and compare semantic segmentation networks that label drivable area and obstacles on unstructured Indian roads using IDD, measuring mIoU, boundary accuracy and speed. The study tests which architectures and training choices cope with missing markings and dense traffic.
9 weeksPython · PyTorch · IDD dataset - M.TechCamera perceptionAdvanced
7.Camera Calibration and Bird's-Eye-View Transform for Lane and Parking Perception
Calibrate a camera's intrinsics and extrinsics in MATLAB and OpenCV, build a metric bird's-eye-view transform of the road and use it for lane and parking-slot measurement. BEV accuracy is checked against surveyed floor markings and its sensitivity to pitch and height errors is quantified.
8 weeksMATLAB · Computer Vision Toolbox · OpenCV - M.TechCamera perceptionAdvanced
8.Real-Time Perception on an Embedded GPU: Quantisation and Latency
Deploy a trained object detector on an embedded GPU board, comparing FP32, FP16 and INT8 TensorRT engines built from ONNX for accuracy loss, latency, throughput and power. The study profiles the full pipeline from camera capture to output and checks it against a real-time ADAS budget.
9 weeksPython · TensorRT · ONNX - PhDCamera perceptionResearch
9.Perception Robustness under Rain, Fog, Night and Glare
Build a benchmark of real (BDD100K, nuScenes) and calibrated synthetic rain, fog, night and glare images, measure how camera object detection degrades in each, and compare robustness remedies. Results are linked to SOTIF triggering conditions.
16 weeksPython · PyTorch - PhDCamera perceptionResearch
10.Uncertainty-Aware Object Detection for Safety-Critical ADAS
Compare Monte Carlo dropout, deep ensembles and probabilistic box regression for object detectors on driving data, measuring calibration and whether uncertainty flags errors and unfamiliar objects. A downstream warning test shows if uncertainty improves safety decisions.
16 weeksPython · PyTorch - B.Tech / BERadar & lidarIntermediate
11.FMCW Radar Signal Processing: Range–Doppler Maps and CFAR Detection
Simulate an automotive FMCW radar in MATLAB, generate beat signals for moving targets, form range–Doppler maps with two-dimensional FFTs and detect targets with CA-CFAR. Estimated range and speed are compared with truth and the effect of waveform parameters on resolution is measured.
6 weeksMATLAB · Radar Toolbox - B.Tech / BERadar & lidarBeginner
12.Ultrasonic Parking Sensor System with Distance Zones and Alerts
Build a reverse parking aid on the DIYguru EV Embedded Development Kit with an HC-SR04 add-on: timed echo measurement on the STM32, filtered distance, zones on the LCD and RGB LEDs, a buzzer that beeps faster as obstacles near, and distance and zone sent on CAN.
4 weeksMicrocontroller · Embedded C - B.Tech / BERadar & lidarIntermediate
13.Lidar Point-Cloud Ground Removal and Obstacle Clustering
Process KITTI lidar scans in Python with Open3D: crop and downsample the cloud, remove the ground with RANSAC plane fitting and a patch-wise variant, cluster the remaining points into obstacles with DBSCAN and fit bounding boxes. Ground removal and obstacle detection are scored against KITTI labels.
6 weeksPython · Open3D - B.Tech / BERadar & lidarIntermediate
14.Radar Target and Clutter Simulation for an ACC Sensor
Model a forward ACC radar in MATLAB: build scenarios in the Driving Scenario Designer, attach a statistical radar model and a waveform-level model, add guardrails and clutter, and compare detections, false alarms and accuracy across Monte Carlo runs.
5 weeksMATLAB · Radar Toolbox - M.TechRadar & lidarAdvanced
15.3-D Object Detection from Lidar with PointPillars
Train and evaluate a PointPillars 3-D detector on KITTI lidar in PyTorch, measuring 3-D and bird's-eye AP for cars, pedestrians and cyclists. Ablations of pillar size, point features and augmentation, plus latency profiling, show what drives accuracy and speed.
10 weeksPython · PyTorch · KITTI dataset - M.TechRadar & lidarAdvanced
16.MIMO Radar Angle-of-Arrival Estimation for Automotive Targets
Simulate a 77 GHz TDM-MIMO automotive radar in MATLAB, form a virtual array and estimate target angles with FFT beamforming, MVDR and MUSIC. Accuracy, resolution of close targets and robustness to SNR, snapshots and calibration errors are measured and compared.
9 weeksMATLAB · Radar Toolbox · Phased Array System Toolbox - M.TechRadar & lidarAdvanced
17.Lidar–Camera Extrinsic Calibration and Projection
Estimate the lidar-to-camera transform from checkerboard data with Lidar Toolbox, validate it by projecting points onto held-out images, and compare it with a targetless edge-based refinement. The study also measures how calibration error affects fused object depth.
8 weeksMATLAB · Lidar Toolbox - PhDRadar & lidarResearch
18.Automotive Radar Interference Detection and Mitigation
Model mutual interference between automotive FMCW radars in MATLAB traffic scenarios, measure how it masks weak targets, and compare mitigation by time-domain zeroing, interpolation, waveform randomisation and a learned denoiser on detection and cost.
16 weeksMATLAB · Radar Toolbox - PhDRadar & lidarResearch
19.4-D Imaging Radar Point Clouds for Object Classification
Develop deep networks that classify and detect objects from 4-D imaging radar point clouds, comparing point-based and pillar-based designs and measuring what Doppler, RCS, elevation and multi-frame accumulation contribute, with a matched comparison against lidar.
18 weeksPython · PyTorch - PhDRadar & lidarResearch
20.Lidar SLAM and Map-Based Localisation for Automated Driving
Build a ROS 2 lidar SLAM and localisation system: ICP or NDT scan registration, a pose graph with loop closure, maps from KITTI odometry sequences and localisation of new drives against them. Drift, accuracy, robustness and runtime are measured against ground truth.
16 weeksROS 2 · Python · C++ - B.Tech / BESensor fusion & trackingIntermediate
21.Kalman Filter Tracking of a Lead Vehicle from Noisy Radar
Build a linear Kalman filter that tracks the range and relative speed of a lead vehicle from noisy simulated radar measurements, in both MATLAB and Python. The filter is tuned, checked for consistency and compared with raw measurements and a moving-average baseline.
5 weeksMATLAB · Python - B.Tech / BESensor fusion & trackingIntermediate
22.Camera–Radar Fusion for Forward Collision Warning
Build a forward collision warning chain in MATLAB that fuses simulated camera and radar detections with a multi-object tracker, selects the most important object in the ego lane and raises a warning from time-to-collision. Fused warnings are compared with camera-only and radar-only versions.
6 weeksMATLAB · Automated Driving Toolbox - B.Tech / BESensor fusion & trackingAdvanced
23.IMU–GNSS Fusion for Vehicle Localisation with an EKF
Build an extended Kalman filter in MATLAB that fuses simulated IMU and GNSS data to estimate a car's position, heading and speed, including through GNSS outages such as tunnels and flyovers. The fused estimate is compared with GNSS alone and with pure dead reckoning.
6 weeksMATLAB · Sensor Fusion and Tracking Toolbox - M.TechSensor fusion & trackingAdvanced
24.Multi-Object Tracking with Global Nearest Neighbour Association
Build and evaluate a multi-object tracker for highway and urban traffic that uses global nearest neighbour data association, gating and M-of-N track management. Performance is measured with GOSPA, track-swap and false-track metrics and compared with a greedy nearest-neighbour baseline.
9 weeksMATLAB · Sensor Fusion and Tracking Toolbox - M.TechSensor fusion & trackingAdvanced
25.JPDA Tracking in Clutter: Comparison with GNN
Build a joint probabilistic data association (JPDA) tracker and compare it with a global nearest neighbour tracker on radar scenarios with increasing clutter and closely spaced vehicles. The study measures GOSPA, track swaps, false tracks and run time to show when the extra computation of JPDA pays off.
9 weeksMATLAB · Sensor Fusion and Tracking Toolbox - M.TechSensor fusion & trackingAdvanced
26.Track-to-Track Fusion of Camera and Radar Trackers
Build a two-level tracking architecture in MATLAB where a camera tracker and a radar tracker each produce tracks that a track-to-track fuser combines. Fusion with cross-covariance and covariance intersection is compared with central detection-level fusion for accuracy, consistency and communication load.
8 weeksMATLAB · Sensor Fusion and Tracking Toolbox - M.TechSensor fusion & trackingAdvanced
27.IMM Filter for Manoeuvring Vehicles (Constant Velocity / Turn Models)
Build an interacting multiple model (IMM) filter that switches between constant-velocity, constant-acceleration and constant-turn models to track vehicles that brake, change lanes and turn. Its accuracy and responsiveness are compared with single-model filters on scripted manoeuvres.
8 weeksMATLAB · Sensor Fusion and Tracking Toolbox - PhDSensor fusion & trackingResearch
28.Learned Bird's-Eye-View Fusion of Camera and Radar
Train and analyse a learned bird's-eye-view (BEV) 3D object detector that fuses surround cameras and radar on the nuScenes dataset. The study measures how radar improves detection and velocity estimation over a camera-only baseline, especially at range, at night and in rain.
18 weeksPython · PyTorch · nuScenes dataset - PhDSensor fusion & trackingResearch
29.Fault-Tolerant Sensor Fusion under Sensor Degradation
Build a camera–radar–lidar fusion system in MATLAB that detects when a sensor is degraded or failing, by blinding, rain, misalignment or dropout, and adapts so tracking stays safe. Degradation detection and fused tracking quality are measured on scripted fault scenarios.
16 weeksMATLAB · Sensor Fusion and Tracking Toolbox - PhDSensor fusion & trackingResearch
30.Cooperative Perception with V2X-Shared Object Lists
Build a cooperative perception system in CARLA where connected vehicles and a roadside unit share object lists over simulated V2X, and the ego vehicle fuses them with its own detections. It measures how sharing improves detection of occluded objects under latency, packet loss and pose error.
16 weeksPython · CARLA - B.Tech / BEACC, AEB & longitudinalIntermediate
31.Adaptive Cruise Control with Spacing and Speed Modes in Simulink
Build an adaptive cruise control function in Simulink and Stateflow that switches between speed control and spacing control with a constant time-gap policy, driving a longitudinal EV model. It is tested on lead-vehicle scenarios for spacing error, comfort and minimum time gap.
5 weeksSimulink · Stateflow - B.Tech / BEACC, AEB & longitudinalBeginner
32.Forward Collision Warning Using Time-to-Collision
Build a forward collision warning function in MATLAB and Simulink that computes time-to-collision and enhanced TTC from lead-vehicle range data and issues staged alerts. Warning timing and false alarms are measured on rear-end scenarios at different speeds and decelerations.
4 weeksMATLAB · Simulink - B.Tech / BEACC, AEB & longitudinalIntermediate
33.Autonomous Emergency Braking with Staged Warnings and Braking
Build an autonomous emergency braking function in Simulink that moves from a warning to partial braking and then full braking based on time-to-collision, using fused camera and radar tracks. It is tested on Euro NCAP-style rear-end scenarios for collision avoidance and impact speed reduction.
5 weeksSimulink · Automated Driving Toolbox - B.Tech / BEACC, AEB & longitudinalIntermediate
34.Stop-and-Go Traffic Jam Assist for Indian City Traffic
Build a stop-and-go traffic jam assist in Simulink and Stateflow that follows a lead vehicle down to standstill, holds the car stopped and resumes smoothly, tuned for dense Indian city traffic. It is tested for gap keeping, comfort, cut-in response and correct standstill behaviour.
6 weeksSimulink · Stateflow - M.TechACC, AEB & longitudinalAdvanced
35.Energy-Aware ACC for EVs with Regenerative Braking
Design an energy-aware adaptive cruise control for an EV that uses model predictive control to trade small spacing variations for more regenerative braking and less wasted energy. It is compared with a conventional ACC on Indian drive-cycle-based traffic for energy use, safety margins and comfort.
9 weeksSimulink · Model Predictive Control Toolbox - M.TechACC, AEB & longitudinalAdvanced
36.MPC-Based Adaptive Cruise Control with Comfort Constraints
Design an MPC-based adaptive cruise control in Simulink with explicit constraints on safe gap, acceleration and jerk, and compare it with a PID-based ACC and the toolbox ACC block. The study measures gap tracking, comfort, constraint violations and solver time on lead-vehicle and cut-in scenarios.
8 weeksSimulink · Model Predictive Control Toolbox - M.TechACC, AEB & longitudinalAdvanced
37.AEB for Pedestrians and Two-Wheelers: Scenario Design and Tuning
Design and tune an AEB function for pedestrians, cyclists and motorcyclists in Simulink, with crossing, longitudinal and turning scenarios built in Driving Scenario Designer and RoadRunner to reflect Indian streets. Performance is measured as avoidance rate, impact speed reduction and false activations.
9 weeksSimulink · Automated Driving Toolbox - PhDACC, AEB & longitudinalResearch
38.Cooperative ACC and Platooning with V2V Communication
Design cooperative adaptive cruise control for a platoon of vehicles that share acceleration over V2V, allowing shorter time gaps than ACC while staying string stable. Controllers are analysed in MATLAB and Simulink and tested in co-simulation with realistic communication delay and packet loss.
16 weeksMATLAB · Simulink - PhDACC, AEB & longitudinalResearch
39.Learning Human-Like Car-Following Behaviour for ACC
Learn human car-following behaviour from naturalistic driving data and driving-simulator sessions, and use it to build an ACC that feels natural while keeping hard safety limits. Learned models are compared with classical car-following models and a conventional ACC on accuracy, safety and comfort.
16 weeksPython · MATLAB - PhDACC, AEB & longitudinalResearch
40.String Stability of ACC Vehicles in Mixed Indian Traffic
Analyse and simulate string stability of ACC vehicles in mixed traffic with human-driven cars, buses and two-wheelers typical of Indian roads. The study finds how ACC penetration rate, time gap, vehicle mix and cut-ins affect the growth of stop-and-go waves, and proposes ACC tuning that damps them.
14 weeksMATLAB · Simulink - B.Tech / BELane keeping & lateralBeginner
41.Lane Departure Warning System with Camera Lane Detection
Build a camera-based lane departure warning system in Python and OpenCV that detects lane lines, estimates the vehicle's lateral offset and time to line crossing, and raises a warning when the car drifts without signalling. It is tested on recorded dashcam footage.
4 weeksPython · OpenCV - B.Tech / BELane keeping & lateralIntermediate
42.Lane Keeping Assist with a PID Steering Controller
Build a lane keeping assist (LKA) function in Simulink that steers a bicycle-model vehicle back to the lane centre with a PID controller on lateral offset and heading error. It is tested on straight and curved roads built in the Driving Scenario Designer.
5 weeksSimulink · Automated Driving Toolbox - B.Tech / BELane keeping & lateralIntermediate
43.Path Tracking: Pure Pursuit vs Stanley Controller
Implement the pure pursuit and Stanley path-tracking controllers in MATLAB or Python on a kinematic bicycle model and compare their cross-track error, stability and steering effort on a common set of reference paths and speeds.
5 weeksMATLAB · Python - B.Tech / BELane keeping & lateralIntermediate
44.Blind-Spot Detection and Lane-Change Warning
Build a blind-spot detection and lane-change warning function in MATLAB that uses simulated rear-corner radars to track vehicles in adjacent lanes and warns when a lane change would be unsafe. It is tested on scripted overtaking and merging scenarios.
5 weeksMATLAB · Automated Driving Toolbox - M.TechLane keeping & lateralAdvanced
45.Lane Keeping with Model Predictive Control
Design a lane keeping controller using model predictive control on a dynamic bicycle model in Simulink, with explicit steering limits and curvature preview, and compare it with a tuned PID baseline across speeds, curvatures and sensor noise.
8 weeksSimulink · Model Predictive Control Toolbox - M.TechLane keeping & lateralAdvanced
46.Highway Lane-Change Planner with Collision Checking
Build a highway lane-change planner in MATLAB that generates candidate trajectories, checks them for collisions against predicted surrounding vehicles and selects the safest comfortable option. It is tested in Automated Driving Toolbox scenarios with merging and slowing traffic.
9 weeksMATLAB · Automated Driving Toolbox - M.TechLane keeping & lateralAdvanced
47.Lateral Assistance Concepts for Two-Wheelers: Lean-Angle-Aware Warnings
Build lean-angle-aware lateral warnings for electric two-wheelers: a Simulink motorcycle model computes safe lean limits for curves, and a bench prototype on the DIYguru EV kit estimates lean from the MPU6050 and warns of excessive lean or a fast curve approach.
8 weeksMATLAB · Simulink - PhDLane keeping & lateralResearch
48.Shared Steering Control between Driver and Automation
Develop and evaluate a shared steering controller in which an automation system and a human driver both apply torque to the steering wheel, with authority that adapts to driver intent and risk. It is studied in Simulink with driver models and a driver-in-the-loop simulator.
16 weeksMATLAB · Simulink - PhDLane keeping & lateralResearch
49.Lane Keeping on Poorly Marked Roads Using Road-Edge Cues
Develop a perception method that estimates a drivable corridor and virtual lane on poorly marked roads from road edges, kerbs and traffic cues, trained and tested on the India Driving Dataset, and feed it to a lane keeping controller in simulation.
16 weeksPython · PyTorch - PhDLane keeping & lateralResearch
50.End-to-End Steering Prediction and Its Safety Envelope
Train an end-to-end neural network that predicts steering from camera images in CARLA, then define and enforce a safety envelope around it with a runtime monitor and a fallback controller, measuring how often and how well the envelope intervenes.
18 weeksPython · PyTorch · CARLA - B.Tech / BEParking & path planningIntermediate
51.Automated Parallel Parking Assist with Ultrasonic Sensing
Build a parallel parking assist that measures a parking gap with side ultrasonic sensors, plans a two-arc reverse manoeuvre and steers a simulated car into the slot in Simulink and Stateflow. A bench prototype on the DIYguru EV kit reads HC-SR04 sensors and gives distance warnings.
6 weeksMATLAB · Simulink - B.Tech / BEParking & path planningBeginner
52.A* Path Planning on an Occupancy Grid for Parking Lots
Implement the A* search algorithm on an occupancy grid of a parking lot in Python or MATLAB, compare heuristics and connectivity, smooth the resulting path and check it against vehicle size with an inflated map. Results are compared with Dijkstra's algorithm.
4 weeksPython · MATLAB - B.Tech / BEParking & path planningIntermediate
53.Rear-View Camera with Dynamic Parking Guidelines
Build a rear-view camera system in Python and OpenCV that calibrates the camera, corrects lens distortion and overlays static distance lines and dynamic trajectory guidelines that bend with the steering angle, with accuracy checked against floor markings.
5 weeksPython · OpenCV - M.TechParking & path planningAdvanced
54.Hybrid A* Planning for Automated Valet Parking
Build an automated valet parking planner in MATLAB with Navigation Toolbox's Hybrid A* on a car park costmap, smooth and track the path, and evaluate success rate, path quality and planning time for perpendicular, angled and parallel bays.
9 weeksMATLAB · Navigation Toolbox - M.TechParking & path planningAdvanced
55.RRT* Motion Planning with Vehicle Kinematic Constraints
Implement RRT* motion planning with Dubins and Reeds-Shepp vehicle constraints in MATLAB or Python, and study how sampling strategy, steering function and rewiring radius affect path cost, convergence and planning time in parking and narrow-passage scenarios.
8 weeksMATLAB · Python - M.TechParking & path planningAdvanced
56.Frenet-Frame Trajectory Planning for Urban Driving
Implement a Frenet-frame trajectory planner for urban driving that samples lateral and longitudinal polynomials, checks them against kinematic limits and predicted obstacles, and selects the lowest-cost trajectory each cycle. It is evaluated in scenarios with stopped vehicles, cut-ins and curves.
9 weeksPython · MATLAB - B.Tech / BEParking & path planningIntermediate
57.Behaviour Planning with a Hierarchical State Machine for Urban Scenarios
Design a hierarchical behaviour planner in Stateflow that chooses between lane following, car following, stopping at junctions, yielding to pedestrians and overtaking in urban scenarios. It is tested in MATLAB driving scenarios with requirement-based test cases.
5 weeksStateflow · MATLAB - PhDParking & path planningResearch
58.Interaction-Aware Motion Planning with Predicted Road-User Trajectories
Develop a motion planner that uses learned, multi-modal predictions of surrounding road users and models how they react to the ego vehicle, then compare it with prediction-only planning on merging and unprotected-turn scenarios built from nuScenes and Waymo Open Dataset scenes.
18 weeksPython · PyTorch - PhDParking & path planningResearch
59.Trajectory Prediction of Mixed Road Users in Dense Indian Traffic
Build and evaluate a trajectory prediction model for mixed road users in dense Indian traffic, including two-wheelers, auto-rickshaws and pedestrians, using a dataset assembled from recorded video with tracking, and compare it with physics-based and learned baselines.
16 weeksPython · PyTorch - PhDParking & path planningResearch
60.Risk-Aware Planning under Perception Uncertainty
Develop a motion planner that propagates perception uncertainty, from detection confidence, position covariance and occlusion, into a quantified collision risk and constrains planning by a risk bound. It is compared with deterministic and worst-case planners in simulated urban scenarios.
16 weeksPython · MATLAB - B.Tech / BEScenario testing & validationBeginner
61.ADAS Test Suite with the Driving Scenario Designer
Build a reusable suite of ADAS test scenarios in the Driving Scenario Designer, from stationary-target braking to cut-ins and pedestrian crossings, and run a simple forward collision warning function against it with automatic pass/fail checks.
4 weeksMATLAB · Automated Driving Toolbox - B.Tech / BEScenario testing & validationIntermediate
62.Testing an ADAS Function in the CARLA Simulator
Implement a radar-based automatic emergency braking (AEB) function in the CARLA simulator with its Python API, run it in synchronous mode against scripted lead-vehicle and pedestrian scenarios, and measure stopping margins and collision rate.
6 weeksCARLA · Python - B.Tech / BEScenario testing & validationIntermediate
63.Building Indian Road Scenes in RoadRunner for ADAS Testing
Build Indian road scenes in RoadRunner, with unmarked lanes, speed breakers, roundabouts, narrow junctions and mixed traffic, export them as ASAM OpenDRIVE, and use them to test a lane detection or braking function in MATLAB and Simulink.
5 weeksRoadRunner · MATLAB - B.Tech / BEScenario testing & validationIntermediate
64.Simulating NCAP-Style AEB Test Scenarios and Scoring
Simulate NCAP-style AEB test scenarios, car-to-car rear stationary, moving and braking plus pedestrian crossings, in MATLAB, run an AEB controller through them and compute a protocol-style score from impact-speed reduction and avoidance.
5 weeksMATLAB · Automated Driving Toolbox - M.TechScenario testing & validationAdvanced
65.Scenario-Based Test Automation with ASAM OpenSCENARIO
Build an automated test pipeline that runs ADAS test cases written in ASAM OpenSCENARIO on CARLA through ScenarioRunner, varies their parameters, evaluates pass criteria and produces regression reports on every software change.
9 weeksCARLA · Python · OpenSCENARIO - M.TechScenario testing & validationAdvanced
66.SIL and HIL Validation of an ADAS Controller
Validate an adaptive cruise control controller through model-, software- and processor-in-the-loop on the STM32 kit, then a bench HIL setup where the controller runs on the board and exchanges CAN messages with a real-time plant and scenario model on the PC.
10 weeksSimulink · Simulink Test · Embedded Coder - M.TechScenario testing & validationAdvanced
67.SOTIF Analysis: Finding Triggering Conditions for a Perception Function
Apply ISO 21448 (SOTIF) to a camera pedestrian detector: vary weather, lighting, occlusion and distance systematically in CARLA, find the triggering conditions that cause missed detections, confirm them on real images and propose measures.
9 weeksMATLAB · CARLA - PhDScenario testing & validationResearch
68.Automated Generation of Critical Test Scenarios by Search and Falsification
Develop and compare search-based methods, Bayesian optimisation, evolutionary search and temporal-logic falsification, that automatically find critical scenarios for an ADAS function, and measure how many distinct failures each finds per simulation budget.
16 weeksMATLAB · Python · CARLA - PhDScenario testing & validationResearch
69.Coverage Metrics for Scenario-Based ADAS Validation
Define and evaluate coverage metrics for scenario-based ADAS validation, from parameter-space and combinatorial coverage to ODD and situation coverage, and test whether higher coverage actually predicts a test suite's ability to find injected faults.
14 weeksPython · OpenSCENARIO - PhDScenario testing & validationResearch
70.Measuring the Sim-to-Real Gap in Perception Testing
Quantify the sim-to-real gap for camera object detection: train and test detectors on CARLA and real datasets such as KITTI, BDD100K and IDD, measure accuracy drop and per-condition ranking agreement, and test methods that make simulated testing more predictive.
16 weeksPython · CARLA · PyTorch - B.Tech / BEAUTOSAR ClassicIntermediate
71.Designing an AUTOSAR Software Component in Simulink with AUTOSAR Blockset
Design an AUTOSAR Classic software component for an EV coolant pump controller in Simulink, map it to AUTOSAR ports, interfaces and runnables with AUTOSAR Blockset, generate ARXML and C code with Embedded Coder, and verify it with back-to-back tests.
5 weeksSimulink · AUTOSAR Blockset · Embedded Coder - B.Tech / BEAUTOSAR ClassicBeginner
72.AUTOSAR Interfaces, Ports and Runnables: Modelling a Window-Lift ECU
Model a car window-lift ECU as AUTOSAR Classic software components with sender-receiver and client-server interfaces, ports, runnables and events, generate the ARXML, and read and check its structure with a small Python script.
4 weeksAUTOSAR Blockset · ARXML - B.Tech / BEAUTOSAR ClassicIntermediate
73.AUTOSAR COM Stack: Mapping Signals to CAN PDUs
Map application signals to CAN PDUs following the AUTOSAR COM stack layering: define signals, packing and transmission modes, implement a COM-PduR-CanIf-like path in C on the STM32 kit, and verify frames, timing and timeouts with a USB-CAN adapter.
5 weeksAUTOSAR Blockset · CAN - B.Tech / BEAUTOSAR ClassicIntermediate
74.AUTOSAR Composition for an EV Battery Management Function
Build an AUTOSAR Classic composition for an EV battery management function, cell monitoring, SOC estimation, protection and contactor control components, in an AUTOSAR architecture model, simulate it with a battery plant and export the composition ARXML.
6 weeksSimulink · AUTOSAR Blockset - B.Tech / BEAUTOSAR ClassicAdvanced
75.AUTOSAR Mode Management: ECU State and Communication Modes
Model AUTOSAR mode management for an EV ECU: EcuM start-up, run and shutdown states, ComM communication modes and BswM-style rules in Stateflow, connect application components through mode-switch interfaces and test wake-up, sleep and shutdown sequences.
6 weeksAUTOSAR Blockset · Stateflow - M.TechAUTOSAR ClassicAdvanced
76.AUTOSAR Diagnostics: DEM Event Handling and DCM UDS Services
Implement AUTOSAR-style diagnostics on the STM32 kit: DEM event handling with debouncing, DTC status bits, operation cycles and freeze frames, and a DCM serving ISO 14229 UDS services over CAN, tested from a Python UDS tester.
9 weeksAUTOSAR Blockset · UDS - M.TechAUTOSAR ClassicAdvanced
77.AUTOSAR Memory Stack: NvM Blocks, Redundancy and Write Strategies
Implement an AUTOSAR-style memory stack on the STM32 kit, NvM block management with native, redundant and dataset blocks above a flash EEPROM emulation layer, and measure data integrity under resets during writes, write latency and flash wear for different write strategies.
8 weeksAUTOSAR Blockset · Embedded C - M.TechAUTOSAR ClassicAdvanced
78.AUTOSAR OS Scheduling and End-to-End Timing Analysis
Map generated runnables of a multi-rate ADAS controller onto fixed-priority preemptive tasks on the STM32 kit, the scheduling model of AUTOSAR OS, measure execution and response times, compare them with response-time analysis and analyse end-to-end latency along cause-effect chains.
9 weeksAUTOSAR Blockset · Embedded Coder - M.TechAUTOSAR ClassicAdvanced
79.AUTOSAR End-to-End (E2E) Protection for Safety-Relevant Signals
Implement AUTOSAR End-to-End (E2E) protection for a safety-relevant torque request on CAN using the STM32 kit and a PC node: CRC, counter and data ID protection with a receiver state machine, then measure detection of injected corruption, repetition, loss, delay and masquerade faults.
8 weeksAUTOSAR Blockset · CAN - PhDAUTOSAR ClassicResearch
80.Migrating Legacy ECU Code to AUTOSAR Classic: Effort and Overhead Study
Migrate a legacy bare-metal EV ECU firmware on the STM32 kit to an AUTOSAR Classic architecture through three strategies, wrapping, partial refactoring and full re-modelling, and measure effort, code changes, memory, CPU and timing overhead and test results for each.
16 weeksAUTOSAR Blockset · Embedded C - B.Tech / BEAUTOSAR Adaptive & SOAIntermediate
81.SOME/IP Services and Service Discovery with vsomeip
Build a small SOME/IP service network on Linux with COVESA vsomeip: a radar-object provider, a consumer and an event subscriber, discovered at runtime through SOME/IP Service Discovery. Traffic is captured in Wireshark and latency, discovery time and failure recovery are measured.
6 weeksC++ · vsomeip · Linux - B.Tech / BEAUTOSAR Adaptive & SOAIntermediate
82.Service-Oriented Architecture for a Software-Defined Vehicle Feature
Design and build one software-defined vehicle feature, a speed-dependent welcome-and-departure lighting function, as cooperating services on Linux with clear interfaces, versioning and a simple service registry. The design is compared with a signal-based version for coupling and update effort.
5 weeksC++ · Linux - B.Tech / BEAUTOSAR Adaptive & SOAIntermediate
83.Automotive Ethernet Basics: VLANs, Timing and Bandwidth for ADAS Data
Study how ADAS data travels on automotive Ethernet by building a software test network on Linux with VLANs, traffic shaping and generated camera, radar and control streams, then measuring bandwidth, latency and jitter against a calculated budget.
5 weeksWireshark · Linux - M.TechAUTOSAR Adaptive & SOAAdvanced
84.AUTOSAR Adaptive ara::com Proxy and Skeleton for a Perception Service
Implement the AUTOSAR Adaptive ara::com pattern for a camera perception service: a skeleton that publishes detected objects and a proxy that discovers, subscribes and calls methods, on vsomeip. Latency and serialisation cost are measured.
10 weeksC++ · AUTOSAR Adaptive - M.TechAUTOSAR Adaptive & SOAAdvanced
85.Adaptive Platform Execution and State Management for an ADAS Application
Build a model of AUTOSAR Adaptive Execution and State Management on Linux: manifests describe ADAS processes, a supervisor starts them in order, and a state manager switches modes such as Parking and Driving and reacts to failures.
9 weeksC++ · AUTOSAR Adaptive - M.TechAUTOSAR Adaptive & SOAAdvanced
86.Over-the-Air Update Workflow on an Adaptive Platform (UCM Concepts)
Build an over-the-air update workflow for an Adaptive-style vehicle computer on Linux, based on UCM concepts: signed packages, verification, A/B installation, activation in a safe vehicle state and rollback, tested against attacks and power loss.
9 weeksC++ · Linux - M.TechAUTOSAR Adaptive & SOAAdvanced
87.Classic–Adaptive Gateway: Bridging CAN Signals to SOME/IP Services
Build a gateway that reads CAN signals from the DIYguru EV kit's STM32 and republishes them as SOME/IP services on Linux with vsomeip, with a filtered command path back. Mapping, latency, freshness checks and CAN faults are measured.
9 weeksC++ · vsomeip · CAN - PhDAUTOSAR Adaptive & SOAResearch
88.Zonal E/E Architecture: Wiring, Latency and Compute Trade-offs
Build a parametric model comparing domain and zonal E/E architectures for an electric car, estimating wiring mass, network latency and compute load from component placement and signal lists. Sensitivity studies show where zonal designs gain and lose.
16 weeksMATLAB · Python - PhDAUTOSAR Adaptive & SOAResearch
89.Deterministic Communication for ADAS with Time-Sensitive Networking
Simulate an in-vehicle ADAS Ethernet network in OMNeT++ and INET and compare strict priority, credit-based shaping and time-aware shaping with frame pre-emption. Worst-case latency, jitter and schedule synthesis effort are evaluated.
16 weeksPython · OMNeT++ - PhDAUTOSAR Adaptive & SOAResearch
90.Containerised Vehicle Software: Isolation, Startup Time and Real-Time Limits
Measure what containerising vehicle software costs and gains: run an ADAS-style pipeline of perception, fusion and control tasks on Linux natively, in Docker containers and with real-time tuning, and compare startup time, latency, jitter, isolation under interference and update effort.
16 weeksLinux · Docker · C++ - B.Tech / BEFunctional safety & cybersecurityBeginner
91.Hazard Analysis and Risk Assessment (HARA) for Lane Keeping Assist
Carry out an ISO 26262 Hazard Analysis and Risk Assessment for a Lane Keeping Assist function: define the item, identify malfunctions, build hazardous events across operating situations, rate severity, exposure and controllability, determine ASILs and write safety goals with safe states.
4 weeksISO 26262 · Spreadsheet - B.Tech / BEFunctional safety & cybersecurityIntermediate
92.FMEA and Fault Tree Analysis of an AEB System
Analyse an Autonomous Emergency Braking system with a design FMEA and fault tree analysis: rate failure modes of sensors, controller, network and brakes, build trees for missed and unintended braking, and evaluate safety mechanisms.
5 weeksISO 26262 · Spreadsheet - B.Tech / BEFunctional safety & cybersecurityIntermediate
93.Threat Analysis and Risk Assessment (TARA) for a Connected EV under ISO/SAE 21434
Perform a Threat Analysis and Risk Assessment under ISO/SAE 21434 for a connected EV's telematics and remote functions: define the item, identify assets and damage scenarios, model threats with STRIDE, map attack paths, rate impact and attack feasibility, determine risk and set cybersecurity goals.
5 weeksISO/SAE 21434 · Spreadsheet - B.Tech / BEFunctional safety & cybersecurityIntermediate
94.Functional and Technical Safety Requirements for Adaptive Cruise Control
Develop functional and technical safety requirements for Adaptive Cruise Control from its safety goals, following ISO 26262: functional safety concept, allocation, safety mechanisms with timing, and traceability to Simulink tests in Requirements Toolbox.
5 weeksISO 26262 · Requirements Toolbox - M.TechFunctional safety & cybersecurityAdvanced
95.Safety Mechanisms: Sensor Plausibility Checks and Degraded Modes
Design safety mechanisms for an ADAS longitudinal function in Simulink and Stateflow: range, rate, cross-sensor and model-based plausibility checks with a degraded-mode manager. Detection time, false alarms and coverage are measured by fault injection.
8 weeksSimulink · Stateflow - M.TechFunctional safety & cybersecurityAdvanced
96.ASIL Decomposition and Safety Architecture for an AEB Function
Design a safety architecture for AEB using ASIL decomposition: compare decomposition options, check independence with a dependent failure analysis, model the options in SysML and evaluate them with fault trees and fault-injection simulation.
8 weeksISO 26262 · SysML - M.TechFunctional safety & cybersecurityAdvanced
97.Secure Onboard Communication (SecOC) for ADAS Signals on CAN
Implement AUTOSAR-style Secure Onboard Communication on CAN: the DIYguru EV kit's STM32 sends ADAS signals with a truncated CMAC and freshness value, and a Linux receiver verifies them. Spoofing, replay and tampering attacks are tested, and the cost in bus load, latency and CPU time is measured.
9 weeksAUTOSAR Blockset · CAN - PhDFunctional safety & cybersecurityResearch
98.Adversarial Attacks on Traffic-Sign Perception and Defences
Train a GTSRB traffic-sign classifier, then evaluate FGSM, PGD and patch attacks and defences such as adversarial training and input transforms under adaptive attacks, measuring robust accuracy, clean accuracy and compute cost.
16 weeksPython · PyTorch - PhDFunctional safety & cybersecurityResearch
99.Runtime Safety Monitoring of Machine-Learning Perception Components
Design and evaluate runtime monitors that flag unreliable outputs of a camera object detector, tested on shifted data such as night, rain and Indian roads, and judged by how well they catch errors that affect a downstream AEB function.
18 weeksPython · PyTorch - PhDFunctional safety & cybersecurityResearch
100.Safety and Security Co-Engineering (ISO 26262 + ISO/SAE 21434) for ADAS
Develop and evaluate a combined safety and security method for an ADAS function, linking ISO 26262 HARA and ISO/SAE 21434 TARA in one SysML model, finding attack-induced hazards and conflicts, and comparing it with separate analyses.
16 weeksSysML · ISO 26262 · ISO/SAE 21434
Learn it at DIYguru
The courses behind these projects
These DIYguru courses and programmes teach the tools and the engineering these projects use, with live classes, recorded lessons, projects and a certificate.
- ProgrammeAdvanced Nanodegree Certification Program in ADAS & AUTOSAR
- CourseCertification Course in Advanced Driver Assistance Systems (ADAS) for EVs
- CourseCertification Course in MATLAB - Simulink
- CourseCertification Course in AUTOSAR for EVs: Standards & Software Architecture
- CourseFundamental of AUTOSAR Engineering
- ProgrammeAdvanced Nanodegree Certification Program in Embedded Systems Design & Development
Questions
Frequently asked questions
Which tools do the ADAS and AUTOSAR projects use?
Perception projects use Python with OpenCV and PyTorch, or MATLAB's Computer Vision and Automated Driving toolboxes. Control and fusion projects use MATLAB, Simulink and the Sensor Fusion and Tracking Toolbox; scenario testing uses the Driving Scenario Designer, RoadRunner or the open-source CARLA simulator; AUTOSAR Classic projects use Simulink with AUTOSAR Blockset, and Adaptive projects use C++ with open-source SOME/IP. Each guide lists exactly what it needs.
Can I do an ADAS project without a real car?
Yes, and you should. Every project here runs in simulation, on public datasets or on a bench. Never test driver-assistance functions on public roads.
Where can I learn ADAS and AUTOSAR?
DIYguru's Advanced Nanodegree in ADAS & AUTOSAR, created with I3C–IIT Jammu, covers perception to ECU integration, AUTOSAR Classic and Adaptive, and ADAS design and validation in MATLAB and Simulink. The ADAS and AUTOSAR certification courses cover each area on its own. Every project links the courses that teach it.