Autonomous Driving Projects
Autonomous Vehicle Projects — Perception, Planning & SimulationAutonomous vehicles combine computer vision, sensor fusion, localization, motion planning and control. Final-year projects that implement detection/lane modules, fuse LiDAR with cameras, plan paths in simulation, or run ROS2 stacks produce strong, demonstrable results.
Below are 60+ topics across perception, fusion, planning, SLAM, CARLA simulation and control/ROS2, with tools used in research and industry (CARLA, ROS2, YOLO, OpenCV, PCL, PyTorch).
CARLA
ROS / ROS2
YOLO / Detectron
OpenCV
PyTorch
PCL / Point Clouds
| # | Autonomous Vehicle Project Topic | Tools Used |
|---|---|---|
| 👁️ Perception — Detection · Segmentation · Lanes · Signs | ||
| 01 | PercReal-Time Vehicle and Pedestrian Detection for AV | YOLO / Detectron, OpenCV, KITTI |
| 02 | PercLane Detection and Tracking with Classical + Deep Methods | OpenCV, CNN, CARLA / video |
| 03 | PercSemantic Segmentation of Road Scenes (Cityscapes-style) | DeepLab / SegFormer, PyTorch |
| 04 | PercTraffic Sign and Traffic Light Recognition | YOLO / CNN, custom or GTSRB |
| 05 | PercDrivable Area / Free-Space Segmentation | Seg models, OpenCV post-process |
| 06 | PercMonocular Depth Estimation for Obstacle Awareness | MiDaS / DepthAnything, PyTorch |
| 07 | PercMulti-Object Tracking (MOT) for Vehicles and Pedestrians | SORT / DeepSORT / ByteTrack |
| 08 | Perc3D Object Detection from Camera (Pseudo-LiDAR Concepts) | Depth + 2D detect, KITTI eval |
| 09 | PercNight / Adverse Weather Detection Robustness Study | Augmentation, YOLO, metrics |
| 10 | PercBird’s-Eye-View (BEV) Representation from Cameras | IPM / BEV models, PyTorch |
| 📡 Sensor Fusion — LiDAR · Camera · Radar | ||
| 11 | FuseLiDAR Point Cloud Object Detection (PointPillars-style) | PCL / Open3D, PyTorch, KITTI |
| 12 | FuseCamera–LiDAR Early / Late Fusion for Detection | Calibration, YOLO + PCL, KITTI |
| 13 | FuseOccupancy Grid Mapping from LiDAR Scans | PCL, grid mapping, ROS optional |
| 14 | FuseSensor Calibration: Camera Intrinsic / Extrinsic + LiDAR | OpenCV, checkerboard, target |
| 15 | FuseMulti-Sensor Tracking with Kalman / UKF Fusion | FilterPy / custom, detections |
| 16 | FuseRadar–Camera Fusion for All-Weather Perception (Concepts) | Simulated / public radar data |
| 17 | FusePoint Cloud Segmentation and Clustering for Obstacles | PCL Euclidean / RANSAC |
| 18 | FuseTemporal Fusion: Multi-Frame Detection Consistency | Tracking, NMS over time |
| 🗺️ Path Planning & Decision Making | ||
| 19 | PlanA* / Hybrid A* Path Planning in Occupancy Grids | Python / C++, grid maps |
| 20 | PlanRRT / RRT* Motion Planning for Non-Holonomic Vehicles | OMPL concepts, custom impl |
| 21 | PlanLattice / Sampling-Based Local Planner | Custom planner, cost maps |
| 22 | PlanBehaviour Planning: Lane Change and Intersection Logic | FSM / behaviour trees, CARLA |
| 23 | PlanModel Predictive Control (MPC) for Trajectory Tracking | CasADi / CVXPY, vehicle model |
| 24 | PlanSpeed Profile Optimisation with Comfort Constraints | Optimisation, longitudinal model |
| 25 | PlanObstacle Avoidance with Dynamic Objects | Prediction + local planner |
| 26 | PlanGlobal Route Planning on OpenStreetMap-style Graphs | NetworkX / osmnx concepts |
| 27 | PlanCostmap Generation from Perception Outputs | Occupancy + inflation layers |
| 📍 SLAM & Localization | ||
| 28 | SLAMVisual Odometry / Visual SLAM Pipeline | OpenCV, ORB-SLAM concepts |
| 29 | SLAMLiDAR Odometry and Mapping (LOAM-style Concepts) | PCL, scan matching |
| 30 | SLAMGNSS + IMU + Wheel Odometry Fusion for Localization | EKF / UKF, ROS localization |
| 31 | SLAMMap Matching and Localization in Prior Maps | NDT / ICP, HD map concepts |
| 32 | SLAMLoop Closure Detection for Mapping Consistency | Bag matching, pose graph |
| 33 | SLAMParticle Filter Localization in Known Maps | AMCL concepts, ROS2 |
| 🖥️ Simulation — CARLA · Scenario Testing | ||
| 34 | SimCARLA-Based Autonomous Driving Agent (Rule or RL) | CARLA, Python API |
| 35 | SimScenario Generation and Evaluation in CARLA | CARLA scenarios, metrics |
| 36 | SimSensor Simulation: Camera / LiDAR Data Collection Pipeline | CARLA sensors, recording |
| 37 | SimClosed-Loop Testing of Perception + Planning Stack | CARLA, ROS2 bridge optional |
| 38 | SimTraffic Scenario: Intersection Handling and Yield Logic | CARLA traffic manager |
| 39 | SimWeather and Lighting Robustness in Simulation | CARLA weather API, detect |
| 40 | SimBenchmark Suite: Success Rate, Collision, Route Completion | CARLA leaderboard-style metrics |
| 🎛️ Control · ROS2 · Systems Integration | ||
| 41 | CtrlPure Pursuit / Stanley Controller for Path Tracking | Python / C++, kinematic model |
| 42 | CtrlLongitudinal Control: PID Speed / ACC Concepts | PID, vehicle dynamics simple |
| 43 | CtrlROS2 Perception Node Pipeline (Camera → Detect → Publish) | ROS2, OpenCV, YOLO |
| 44 | CtrlROS2 Navigation Stack Concepts for Differential / Ackermann | Nav2 concepts, costmaps |
| 45 | CtrlMessage Synchronisation and TF Transforms in ROS2 | ROS2, message_filters, tf2 |
| 46 | CtrlSafety Layer: Collision Check and Emergency Stop Logic | Custom, costmap / distance |
| 47 | CtrlHardware-in-the-Loop Concepts with Embedded Controllers | Jetson / MCU, CAN concepts |
| 🧠 Learning-Based & End-to-End Approaches | ||
| 48 | LearnImitation Learning for Lane Following in CARLA | CARLA, behavioural cloning |
| 49 | LearnEnd-to-End Steering Prediction from Front Camera | CNN, Udacity / CARLA data |
| 50 | LearnReinforcement Learning Agent for Simple AV Tasks | CARLA RL, Stable-Baselines |
| 51 | LearnDomain Adaptation: Sim-to-Real for Detection Models | CARLA → real images, DA methods |
| 52 | LearnUncertainty Estimation in Perception Outputs | MC Dropout / ensembles, YOLO |
| 📊 Datasets · Evaluation · Safety | ||
| 53 | EvalKITTI / nuScenes Evaluation Pipeline for Detection | Official eval scripts, mAP |
| 54 | EvalLane Detection Metrics (F1, IoU) and Visualisation | Custom metrics, OpenCV |
| 55 | EvalSafety Metrics: Time-to-Collision, Minimum Distance | Trajectory analysis, CARLA |
| 56 | EvalAblation Study: Component Contribution in Modular Stack | Controlled experiments |
| 57 | EvalDataset Curation and Annotation Workflow for AV | Label tools, guidelines |
| 🔬 Applied & Research-Oriented Topics | ||
| 58 | AppliedParking Slot Detection and Automated Parking Concepts | Vision / LiDAR, path planner |
| 59 | AppliedV2X / Infrastructure-Assisted Perception Concepts | Simulated messages, fusion |
| 60 | AppliedHuman–AV Interaction: Intention Prediction for Pedestrians | Pose / trajectory models |
| 61 | ResearchScalable Scenario Testing Framework Design | CARLA, scenario DSL |
| 62 | ResearchFailure Mode Analysis of Perception in Edge Cases | Error taxonomy, datasets |
| 63 | ResearchReproducible AV Experiment Protocol for Student Projects | Configs, seeds, logging |
| 64 | ResearchComparative Study: Camera-Only vs Camera+LiDAR Stacks | Same scenes, metrics |
| 65 | ResearchComfort-Oriented Trajectory Generation Benchmark | Jerk/acceleration metrics |
Topics use widely available tools (CARLA, ROS2, YOLO, OpenCV, PCL) and public datasets (KITTI, Cityscapes concepts). Contact us for reference material, simulation/perception code, evaluation setup, university-format report, PPT and viva Q&A for any topic above.
Why Choose Us for Autonomous Vehicle Projects?
Bangalore-based guidance for BE, BTech and MTech students working on perception, planning and CARLA-based systems.
Perception
Detection, lanes, segmentation and tracking with YOLO, OpenCV and clear metrics on standard datasets.
Sensor Fusion
LiDAR–camera fusion, occupancy grids and multi-sensor tracking with PCL and calibration pipelines.
Planning & Control
A*, RRT, MPC and pure pursuit with costmaps and comfort-aware trajectory design.
CARLA Simulation
Closed-loop agents, scenario testing and sensor data collection in the CARLA simulator.
Frequently Asked Questions — Autonomous Vehicle Projects
Top topics include object and lane detection, LiDAR–camera fusion, path planning (A*, RRT, MPC), SLAM localization, CARLA-based agents and scenarios, ROS2 perception/navigation pipelines, and evaluation with safety metrics.
CARLA simulator, ROS / ROS2, OpenCV, YOLO / Detectron, PyTorch / TensorFlow, Point Cloud Library (PCL), Autoware concepts, Gazebo optional, and datasets such as KITTI and Cityscapes-style data.
Yes. Packages include reference material, simulation or perception code, evaluation metrics, demo notes, university-format report, PPT and viva Q&A covering architecture, sensors and results.
Modular stacks separate perception, localization, planning and control with explicit interfaces. End-to-end learning maps sensors directly to controls. Many student projects focus on modular components (detection, fusion, planning), which are easier to evaluate and debug.
Autonomous Vehicle Project Lab — Bangalore
Simulation, perception and ROS2 guidance for BE, BTech and MTech autonomous vehicle projects.
Detection & Lane
Perception Lab
YOLO / OpenCVPerception Lab
LiDAR Fusion
& PCL
Sensor Fusion& PCL
Path Planning
A* / RRT / MPC
Motion PlanningA* / RRT / MPC
CARLA
Simulation Lab
Closed-Loop AgentsSimulation Lab
ROS2 Perception
& Nav Concepts
Middleware& Nav Concepts
SLAM / Localization
Experiments
Odometry & MapsExperiments
Safety Metrics
& Evaluation
TTC / Success Rate& Evaluation
Viva & Report
Preparation
DocumentationPreparation