Adaptive Path Planning and Collision Avoidance for Autonomous Vehicles on Unstructured Indian Roads
MathWorks · Robotics and Drones · Software
The best-specified statement on the portal with named metrics and linked data, but it is also one of the most attractive and the scope is genuinely massive, so only take it if your team already knows MATLAB and RoadRunner and can spend the whole time on the prediction model rather than the tooling.
Data: Built-in sensor and scenario datasets in MathWorks Automated Driving Toolbox, RoadRunner sample scenes, team-created synthetic scenarios, and public Indian traffic datasets. Indian Driving Dataset (IDD): https://idd.insaan.iiit.ac.in/ ; Mendeley traffic data.
What it actually is
Self-driving systems are built for roads with painted lanes, working signals and traffic that behaves predictably, and Indian roads are none of those things. Cars, auto-rickshaws, bicycles, pedestrians and cattle share the same space and move without warning. The ask is a simulated driving system that can plan a safe path through exactly that kind of chaos and replan the instant something changes.
What to build
A closed-loop simulation pipeline in MATLAB and Simulink with a perception stage fusing camera, LiDAR and radar to detect the mixed road users the statement names including auto-rickshaws, pushcarts, pedestrians and animals, a short-horizon motion prediction stage that handles non-lane-based and irregular trajectories rather than assuming lane-following, a planner producing collision-free paths without relying on lane geometry and replanning in real time, decision logic for informal merging and sudden obstacles, and a vehicle dynamics model closing the loop — validated across the five scenarios the statement mandates including an unmarked village road, an unsignalled urban intersection, a highway merge, a dense market and a cattle crossing, with at least two detailed RoadRunner scenes and results reported on replanning latency, path smoothness and scenario completion rate.
Smallest thing that wins the room
Run the cattle-crossing scenario live and show the planned trajectory re-forming around the animal with the replanning latency displayed on screen as it happens.
How crowded this one gets
A guess, projected from the 2025 statements — the last year where both the submission counts and the winners were published.
Quieter than 75% of the 226 · #57 of 226 by expected field
A normal-sized field. Your idea has to be good, not miraculous.
Why: company-sponsored statements drew the smallest fields of all.
This is a guess, not a fact
Nobody has published 2026’s numbers yet. This is an analysed estimate from last year’s pattern, so please do not take it as the truth — check the live counter on the SIH portal before you decide anything. The range covers the middle half of likely outcomes, so one statement in two lands outside it. Entry closes at 500 ideas per statement, so no range goes past that — a statement that reaches the cap fills and shuts rather than drawing an unlimited crowd. The model reads only three things a team can see before choosing — software or hardware, the theme, and what kind of body posted it — and those explain about a quarter of the variation in last year’s field sizes (R² 0.25 on held-out statements). Trust the band more than the number, and the ordering more than either. It cannot see how good your idea is, which is the part that actually decides it.
The scores
The number is the shorthand. The line under it is the reason.
Acceptance potential
3/5The specification, the linked datasets and the prescribed metrics are all in its favour, but autonomous driving is one of the most attractive-sounding statements on the portal so the field will be crowded with strong MATLAB teams, and the scope is genuinely enormous for the time available.
Feasibility
4/5This is entirely simulation so there is no vehicle, no sensor and no track to arrange, the Indian Driving Dataset is public and linked, and the toolchain the statement names is exactly the toolchain the sponsor provides — the only real gate is whether your institution's licence covers Automated Driving Toolbox, RoadRunner and Deep Learning Toolbox.
Innovation scope
3/5The pipeline stages, the five validation scenarios, the reported metrics and even the recommended toolchain are all specified, so your genuine design freedom is concentrated in the planner formulation and the prediction model for non-lane-based motion.
Clarity
5/5Exceptionally well written for a hackathon statement — it names the five required scenarios individually, splits the expected solution into three explicit parts, states the three evaluation metrics, and lists the exact submission artifacts down to the demonstration video.
Effort
MassivePerception, prediction, planning, decision logic and vehicle dynamics integrated into one closed loop, plus two detailed RoadRunner scenes and five validated scenarios, plus a metrics study, technical report and demo video, is several projects stacked together and RoadRunner scene authoring alone has a steep learning curve.
Demo-ability
EasyA simulated vehicle threading through a market or braking for a cow is immediately legible to any judge without explanation, and the required demonstration video means the strongest visual moment is a mandated deliverable rather than an afterthought.
In its favour
- Green flag: The evaluation metrics are named in the statement — replanning latency, path smoothness, scenario completion rate — so you know exactly what you will be measured on, which is rare and enormously valuable
- Green flag: The Indian Driving Dataset is public, linked and genuinely captures the mixed traffic the statement describes, so your detector trains on the actual conditions rather than on European road footage
- Green flag: Pure simulation means no hardware, no track access and no safety approvals, so the whole project is executable from a laptop
- Green flag: MathWorks is a T1 sponsor that supplies the toolchain, provides student licences for the event and knows this problem well, so a technically sound submission is evaluated by people who can actually judge it
Against it
- Red flag: Licence coverage for Automated Driving Toolbox, RoadRunner and Deep Learning Toolbox is not universal on campus installations — confirm you can actually launch these tools before you commit, because the entire statement assumes them
- Red flag: RoadRunner scene authoring is slow and unfamiliar, and two detailed scenes are a hard requirement, so budget days rather than hours for content creation that produces no algorithmic progress
- Red flag: Autonomous driving attracts the strongest and most numerous entries of any theme, so being competent here is not enough — your prediction of irregular non-lane-based motion is the differentiator, not the planner
- Red flag: Simulation-only validation means you never demonstrate real-world performance, and a judge asking how your perception behaves on genuine IDD footage rather than synthetic RoadRunner renders has a fair and difficult question
What you will be writing
- MATLAB Automated Driving Toolbox sensor fusion
- RoadRunner scenario and scene authoring
- Navigation Toolbox with hybrid A* and TEB planning
- Stateflow behavioural decision logic
- Simulink bicycle model vehicle dynamics
- Indian Driving Dataset for detector training
- Autonomous driving
- Motion planning and control
- Simulation and validation
Prior art to read before you start
adaptive path planning in unstructured environments · mixed traffic trajectory prediction · scenario-based autonomous driving validation
Analysed by Claude Opus. Every score above is a judgment call with its reasoning attached — kindly cross-check this against the official statement on the SIH portal before your team commits to it.