FlexWorm: Primitive-augmented Hybrid Contact-motion Planning for Suction-based Multi-segment Deformable Robots
AuthorsZili Tang, Tiecheng Guo, Qinyue Zhang, Meng Guo
Resources
FlexWorm helps soft, suction-powered robots navigate complex surfaces by combining geometric planning with learned motion snippets and fallback search.
Key results
Rollout episodes used to construct the PaHS primitive library
Validated motion primitives remaining after indexing and duplicate removal
Seconds across the 60-task simulation benchmark
Seconds across the same simulation benchmark
PaHS learned-embedding hit rate in seen environments
PaHS learned-embedding hit rate in unseen environments
What the paper found
FlexWorm introduces a hybrid contact-motion planner for suction-based, multi-segment deformable robots navigating complex 3D surfaces without hand-designed gaits. Its core Block-wise IK Hybrid Search, or IKHS, decomposes the robot at adhesion anchors, searches discrete suction-mode transitions, and solves reduced inverse kinematics only for free body blocks while enforcing screw-theoretic deformation, signed-distance-field collision and adhesion constraints, and a gravity-conditioned quasi-static feasibility polytope. Primitive-augmented Hybrid Search, or PaHS, accelerates this process by retrieving four-step validated motion primitives from a learned dual-tower 16-D embedding, then refining them with IK and falling back to IKHS when retrieval fails. A library built from 1500 rollout episodes contained about 6000 primitives. Across 60 simulation tasks, PaHS matched IKHS with 20/20 success in each of three scenarios, while reducing mean planning time from 16.0 seconds to 1.5 seconds and mean execution steps from 55.7 to 43.7 without increasing deformation cost. Learned retrieval achieved 87.9% hit rate in seen environments and 80.5% in unseen environments, outperforming random and pooled-PCA embeddings. Hardware tests on a two-segment pneumatic robot achieved 9/10 planar obstacle-avoidance successes and 8/10 transitions onto a 45-degree slope, with online recovery from adhesion and actuation errors. Experiments ran using IPOPT, CasADi, PyBullet, Open3D, and an NVIDIA RTX 4070.
Original abstract
Multi-segment suction-based soft robots are promising for inspection and maintenance in confined or fragile environments, but existing approaches still depend heavily on manually designed gaits and environment-specific motion scripts. This work presents a planning framework for serial multi-segment soft robots with deformable body segments and boundary suction pads. The formulation targets full 3D navigation on complex surfaces and explicitly handles discrete adhesion switching and continuous body deformation under geometric, collision, and quasi-static feasibility constraints, while remaining agnostic to the specific actuation realization used to produce segment deformation. Its core, block-wise IK hybrid search (IKHS), performs best-first search over feasible adhesion transitions while solving inverse kinematics only on induced free blocks. On top of IKHS, primitive-augmented hybrid search (PaHS) uses a learned observation--primitive embedding to retrieve short validated motion segments for fast local proposal, with fallback to standard IKHS branching when retrieval fails. In simulation, the framework consistently outperforms controlled baselines in planning success, transition quality, and efficiency across diverse terrains. PaHS matches IKHS in success rate while substantially reducing planning time. Repeated hardware experiments on a pneumatic multi-segment soft robot further demonstrate executability and online recovery under actuation and adhesion uncertainty.
Read the original paperMore in Robotics
Browse all 50 papers →JAMB: Joint Action-Motion Diffusion for Bimanual Manipulation
Chuyang Xiao, Peilin Meng, David Held
JAMB helps two robot arms coordinate by jointly imagining their future movements and the changing 3D scene before acting.
Rolling-WAM: World Action Models with Rolling Imagination
Yinghua Zhou, Junjie Ye, Yiqi Zhao, Hao Dong, Celina Shiyu Wang, Ruohai Ge, Tingyi Yang, Basile Van Hoorick, Gaurav Sukhatme, Vitor Guizilini, Yue Wang
Rolling-WAM keeps future robot actions partially imagined and refined over time, making world-model-based manipulation replan 4.5 times faster.
Training-free Behavior Cloning
Maximilian Adang, Timothy Chen, Lars Osterberg, Aiden Swann, Mac Schwager
A fast, training-free robot controller reuses and corrects demonstration trajectories to deliver traceable behavior at real-time speeds.