System design improves with contracts using agent delegation quotient
Contract-Based Agentic Systems Engineering: Delegation as Quotient for Assured System Design
Robotics
Summary
When building complex systems with many parts, a new part can meet its instructions but still cause problems because it misses what other parts expect from it. The authors show how to create a precise contract for this new part by calculating a 'quotient' that captures all assumptions from its neighbors. This method uses formal requirements language and logic to ensure parts fit together properly. They tested this by having an AI control a robot in a warehouse following these contracts and compared different instruction types to see how well it met overall system goals.
What this means in practice
- •For system engineers: Calculate precise component requirements that guarantee system-wide correctness when delegating design tasks to different teams or tools.
- •For robotics developers: Use contract quotients to safely assign local controller design to AI agents while ensuring overall robot system requirements are met.
Authors
Sai Sandeep Damera, Ryan Matheu, John S. Baras
Abstract
A component an agent builds from a prompt can satisfy it and still break the system it joins, because the prompt omits what its neighbours assume of it. The artifact to hand over is computed, not written. A SysML v2 requirement with assumed and required constraints is an assume/guarantee contract, so when a system requirement is decomposed over fixed neighbours, the open component's requirement is the contract quotient, the weakest sound one, and again a SysML v2 requirement. Its required constraints carry the neighbours' assumptions, which a requirement for the mission alone drops and unit tests with mocked neighbours cannot reach. Signal Temporal Logic is closed under the quotient. We build the requirements, with derivation, allocation, satisfaction and verification, as a SysML~v2 model and compute the quotient from it. A case study delegates the local controller of a simulated warehouse robot, whose route crosses a worker's path and ends at a dock, to one language model under the quotient requirement, a mission-only requirement and a natural-language task, and scores the deliveries on live episodes against the system requirement from ground truth.