A chess engine built from classical search algorithms with machine learning evaluation functions. Designed for low-latency real-time play and tight integration with the physical chess cobot system.
Describe the motivation for building a custom engine rather than using an existing one (Stockfish, etc.) — was it about integration, learning, or constraints on the hardware? Give a brief overview of how the engine works at a high level.
Replace this placeholder text with real project details when you're ready to publish.
Describe your implementation of minimax with alpha-beta pruning. What depth did you search to? Did you implement any extensions — iterative deepening, quiescence search, move ordering? What was the node throughput and typical response latency?
Explain the position evaluation approach. What material and positional features did you use? How did you incorporate machine learning — was this a trained neural network, piece-square tables learned from a dataset, or something else? What data did you train on?
Describe the interface between the chess engine and the physical cobot. How did the engine receive the board state, decide on a move, and communicate back? What latency constraints did the robot impose on the engine?
See also: Chess-Playing Cobot project →
Engine architecture / system diagram