LLM chaining improves robot task planning in home environments
Design and Evaluation of LLM Chaining-Based Task Planning for General Purpose Service Robots
RoboticsArtificial IntelligenceComputation and Language
Summary
Robots that help around the house need to understand and follow complex instructions in natural language. The authors found that using a two-step approach—first classifying the instruction type, then generating detailed actions—helps robots plan tasks better than trying to do everything in one step. They tested this method with different language models and on a real robot, showing it plans more consistently. However, completing tasks also depends on how well the robot executes those plans, which still has challenges.
What this means in practice
- •For service robot developers: Improve task planning reliability in domestic robots by using a two-stage LLM-based approach to handle multi-step natural language instructions.
- •For language model implementers: Design LLM workflows that reduce prompt length and context issues by splitting classification and generation, improving planning consistency across models.
Authors
Lucas Da Mota Bruno, Jiahao Sim, Yoshinobu Hagiwara
Abstract
General Purpose Service Robot (GPSR) tasks, as defined in the RoboCup@Home benchmark, require robots to interpret diverse natural language commands and generate multi-step action sequences in real home environments. Conventional Single Prompt (SP) approaches suffer from context bloat and the "Lost in the Middle" phenomenon, leading to unreliable task planning. We propose an LLM chaining architecture that separates instruction classification and action generation into two specialized stages, reducing per-inference prompt length by approximately 45% while improving planning consistency. We evaluate our method using 100 randomly generated GPSR commands across three language models spanning local open-source and frontier cloud deployment contexts. Results show consistent planning improvements over SP across all models, with gains of up to +37 percentage points on local models. Further, real-robot execution experiments on the Toyota Human Support Robot (HSR) reveal that planning success alone does not guarantee task completion, with 6 of 10 tasks completing successfully and execution-layer failures identified as the primary remaining bottleneck.