Self-Learning RL Agent for Supply Chain Management by Kinaxis

The USPTO published Kinaxis Inc.'s patent application US20260099785A1 for a self-learning reinforcement learning (RL) agent used in enterprise production scheduling. The invention enables the RL agent to interact with a simulated production environment modeled as a dynamic graph, allowing efficient handling of multi-stage scheduling dependencies. The system autonomously learns optimal scheduling policies through iterative training, inference, and continuous learning modes while incorporating user preferences. Key components include a data profiler for historical analysis, a synthesizer for training data generation, and an initializer for environment setup.

Manufacturers and supply chain software developers should review the patent claims to assess potential licensing needs or design-around opportunities. The disclosed methodology using dynamic graph modeling and RL-based policy learning for production scheduling represents a potentially significant advancement over traditional heuristics and genetic algorithms. Companies developing competing AI-driven supply chain optimization solutions should evaluate whether their systems fall within the scope of the disclosed claims.