Single-pass policies improve robotic actions with diverse and consistent plans
Conditional Trajectory Peaks: Single-Pass Multimodal Policies over Action Chunks
Robotics
Summary
Making robots perform complex tasks often means planning many possible ways to act and keeping their actions consistent over time. The authors introduce Conditional Trajectory Peaks (CTP), a method that predicts multiple possible action sequences all at once while ensuring these plans stay consistent when updated. Their method performs well in various tests, including real-world tasks where robots handled objects in different ways without delay. This shows that CTP can help robots act flexibly and reliably with faster decision-making.
What this means in practice
- •For robotic system engineers: Improve robot manipulation by generating multiple consistent action plans quickly for tasks like object placement and sorting.
- •For automation platform developers: Build faster decision-making software for dual-arm robots that maintain flexible but reliable behavior during complex tasks.
Authors
Di Wu, Rongtian Shen, Ping Liu, Xuhua Chen, He Zheng, Lingfeng Zhang, Tao Zhang
Abstract
Multimodal imitation learning requires diverse executable futures under the same observation and consistent behavior across replanning cycles. We present Conditional Trajectory Peaks (CTP), a single-pass policy framework that jointly predicts complete action-chunk candidates, probability masses, and trajectory scales. Distribution-Aware Peak Specialization (DAPS) specializes trajectory peaks using trajectory-level posterior responsibilities and mass- and scale-modulated overlap constraints. Evidence-Gated Trajectory Belief Transport (ETBT) maintains cross-chunk consistency through geometric correspondence between exchangeable candidate sets, while allowing current policy evidence to override historical constraints. CTP achieves a coverage score of 91.40% on Push-T; success rates of 100.0%, 79.72%, and 84.44% on D3IL Avoiding, Aligning, and Sorting-2, respectively. On LIBERO, CTP achieves an average success rate of 97.25%. In real-world dual-arm experiments, CTP preserves both placement modes in a two-plate task, succeeding in all 50 trials. On bottle uprighting and pen placement into a holder, it maintains success rates comparable to $π_{0.5}$ while reducing policy inference latency from 218.24 ms to 75.80 ms. These results demonstrate that single-pass trajectory modeling can combine multimodal behavior, closed-loop consistency, and efficient inference.