Mark D. Pendrith

About

Mark D. Pendrith is a researcher whose work lies at the intersection of reinforcement learning, robotics, and situated agent architectures. His most significant contribution is the introduction of the RL-TOPS architecture, a pioneering hybrid system that combines teleo-reactive planning with reinforcement learning to enable modularity and re-use in robot learning. This framework, detailed in his 1998 paper (37 citations), addresses a fundamental challenge in the field: how to speed up learning by decomposing complex tasks into hierarchies of simpler, learnable behaviours. Pendrith’s approach allows autonomous agents to acquire skills more efficiently, bridging the gap between high-level planning and low-level control. In his subsequent 2000 work (8 citations), he further explored the theoretical and practical hurdles of applying reinforcement learning to real-world, situated agents—tackling issues like state-space explosion and delayed rewards. Though his citation counts are modest, Pendrith’s ideas on hierarchical learning and modular design have influenced later developments in scalable robotic learning systems. His research remains a valuable reference for students and engineers seeking to build adaptive, task-decomposing agents in complex environments.

Research Focus

Key Achievements

Top Papers

1. 1
   RL-TOPS: An Architecture for Modularity and Re-Use in Reinforcement Learning

   37 citations · 1998
2. 2
   Reinforcement Learning in Situated Agents: Theoretical Problems and Practical Solutions

   8 citations · 2000

Key Collaborators

Contact & Links