Run a Monte Carlo Tree Search Strategy for a Robotics Pick-and-Place Task
Overview
What this challenge is about.
You receive a simulator of the pick-and-place task: a bin with 10 randomly-placed parts, an action space of which part to pick next, and a reward = parts picked per minute with penalties for failed picks. The greedy baseline picks the topmost part by visual confidence. Implement an MCTS planner with a planning budget (e.g., 200 simulations per decision), benchmark on 500 episodes, and quantify the throughput gain at multiple planning budgets. Recommend a deployment configuration that fits the on-arm compute budget.
The Brief
What you'll do, and what you'll demonstrate.
Quantify whether MCTS planning beats the current greedy pick policy by enough to justify the on-arm compute cost.
Earning criteria — what you'll demonstrate
- Implement MCTS with UCB1 selection and rollout policies
- Reason about the planning-budget vs. solution-quality trade-off
- Benchmark planners fairly with controlled stochasticity
- Translate planner performance into deployment recommendations
Program Fit
Where this fits in your program.
Sharpens the same skills your degree expects you to demonstrate.
Skills
Skills you'll demonstrate.
Each one shows up on your verified credential.
Careers
Roles this prepares you for.
Real titles. Real skill bridges. Pick the one closest to your trajectory.
ML Researcher
MCTS work on a robotics simulator is a strong portfolio piece for RL and planning research roles.
This challenge sharpens
- monte-carlo-tree-search
- planning
- policy-evaluation
Applied AI Scientist
Trading planner quality against on-arm compute is exactly the kind of applied scientist judgment robotics companies hire for.
This challenge sharpens
- monte-carlo-tree-search
- compute-budgeting
- simulation
Machine Learning Engineer
Implementing planners with deterministic seeding is foundational MLE work on robotics or game-AI teams.
This challenge sharpens
- python
- simulation
- policy-evaluation