Edge-AI Based Distributed Fleet Coordination for Autonomous Mobile Robots (AMRs) in Smart Warehouses
Bharat Electronics Limited · Robotics and Drones · Software
The best-defined problem in this block by some distance — the numeric success criteria and explicit acceptance of simulation make it unusually winnable, so implement the baseline honestly and let the measured comparison carry the demo.
What it actually is
Warehouse robots usually get their routes from a central server, which breaks down when the network lags or a Wi-Fi dead zone appears. The ask is a system where robots talk directly to each other and resolve conflicts themselves, running on small onboard computers. Crucially, the expected deliverable is a simulation, and the success criteria are given as hard numbers.
What to build
A decentralised coordination framework for at least three robots where each agent runs its own planner and communicates peer to peer, sharing position and intended path without any central coordinator; a conflict resolution layer that detects impending collisions and deadlocks at narrow aisles and choke points and negotiates priority locally; dynamic task reallocation and rerouting when an aisle becomes blocked; all demonstrated in a multi-robot simulation with the planning stack running on actual edge hardware, plus a lightweight dashboard showing every robot's live position and battery state.
Smallest thing that wins the room
Send three robots on deliberately crossing paths through a single narrow choke point and show them negotiating passage order peer to peer with zero collisions, then display the total completion time against the stop-and-wait baseline showing the required improvement.
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 · #58 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
5/5The standout of this block — the success criteria are given as hard numbers so your result is unarguable, simulation is explicitly accepted so there is no hardware risk, and multi-agent path finding intimidates enough teams that the field will be far thinner than a defence PSU problem deserves.
Feasibility
4/5The Expected Solution explicitly asks for a multi-robot simulation rather than physical robots, which removes the hardware barrier entirely, and decentralised multi-agent path finding has well-documented algorithms you can implement and benchmark.
Innovation scope
4/5The decentralised constraint is set but the negotiation protocol, the deadlock resolution strategy and the task reallocation policy are all yours, and there are several genuinely different valid architectures.
Clarity
5/5The description numbers three required capabilities, names the deliverable components, and states quantitative success criteria — zero inter-robot collisions and at least a twenty percent reduction in task completion time against stop-and-wait — which is the clearest acceptance bar in this entire block.
Effort
HeavyThe simulation environment, the peer-to-peer messaging layer, the distributed planner and the dashboard are four pieces, though the simulation removes all hardware integration work.
Demo-ability
EasyRobots visibly negotiating a choke point without colliding, with a live counter proving the time saved against the baseline, is self-explanatory and quantitative at the same time.
In its favour
- Green flag: The stated success criteria are numeric, so you know exactly what winning looks like and can report it as a measured result rather than a claim
- Green flag: Simulation is named in the Expected Solution, which removes the robot-purchasing barrier that makes most robotics statements unapproachable
- Green flag: The stop-and-wait baseline is specified for you, so your comparison is pre-defined and cannot be accused of being cherry-picked
- Green flag: Running the identical planner stack on three Raspberry Pis driving simulated robots satisfies the edge-hardware wording cheaply and convincingly
Against it
- Red flag: The description says the system must run on edge hardware while the deliverable is a simulation, so resolve that tension explicitly — run the real stack on real Pis against a simulated world rather than simulating the compute too
- Red flag: Truly decentralised conflict resolution is much harder than centralised planning, and a team that quietly runs a central planner and calls it distributed will be caught by any robotics judge
- Red flag: Deadlock at a choke point is the specific hard case named in the description, so a demo that only shows robots avoiding each other in open space has dodged the actual problem
- Red flag: The twenty percent improvement must be measured against a fairly implemented baseline — a deliberately weak stop-and-wait implementation makes the number meaningless
What you will be writing
- Conflict-Based Search / Priority-Based Search MAPF
- ROS 2 with DDS peer-to-peer discovery
- Gazebo or custom grid simulation
- Raspberry Pi / Jetson Nano edge deployment
- Deadlock detection and local negotiation protocol
- React or Streamlit fleet dashboard
- Multi-agent systems
- Warehouse robotics
- Distributed coordination
Prior art to read before you start
decentralised multi-robot path planning · deadlock resolution at choke points · edge-based fleet coordination
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.