REVIEW 5 cited by
My Body is a Cage: the Role of Morphology in Graph-Based Incompatible Control
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
read the original abstract
Multitask Reinforcement Learning is a promising way to obtain models with better performance, generalisation, data efficiency, and robustness. Most existing work is limited to compatible settings, where the state and action space dimensions are the same across tasks. Graph Neural Networks (GNN) are one way to address incompatible environments, because they can process graphs of arbitrary size. They also allow practitioners to inject biases encoded in the structure of the input graph. Existing work in graph-based continuous control uses the physical morphology of the agent to construct the input graph, i.e., encoding limb features as node labels and using edges to connect the nodes if their corresponded limbs are physically connected. In this work, we present a series of ablations on existing methods that show that morphological information encoded in the graph does not improve their performance. Motivated by the hypothesis that any benefits GNNs extract from the graph structure are outweighed by difficulties they create for message passing, we also propose Amorpheus, a transformer-based approach. Further results show that, while Amorpheus ignores the morphological information that GNNs encode, it nonetheless substantially outperforms GNN-based methods that use the morphological information to define the message-passing scheme.
Forward citations
Cited by 5 Pith papers
-
Transformer Transformer: A Unified Model for Motion-Conditioned Robot Co-design
A single diffusion transformer trains on tokenized robot bodies and motions to generate and optimize robot designs for unseen rewards and trajectories, outpacing evolutionary search in speed and often in reward.
-
Building2Building: A Large Scale Benchmark for Generalizable Real-World Reinforcement Learning
A 6,000-environment EnergyPlus benchmark with heterogeneous observation and action spaces for studying generalization and transfer in RL-based HVAC control.
-
UMI-on-Air: Embodiment-Aware Guidance for Embodiment-Agnostic Visuomotor Policies
Embodiment-Aware Diffusion Policy steers a UMI-trained diffusion policy with controller tracking-cost gradients at inference time, improving aerial manipulation success in simulation and real flights.
-
UMC: Unified Resilient Controller for Legged Robots with Joint Malfunctions
UMC uses masked attention and two-stage training so one policy keeps legged robots walking under eight sensor and joint failure scenarios, cutting fall rates by up to 53 percentage points in simulation.
-
UniLegs: Universal Multi-Legged Robot Control through Morphology-Agnostic Policy Distillation
A two-stage teacher-student distillation produces a single Transformer policy that reaches 94.47% of specialist teacher reward on five training morphologies and 72.64% on an unseen quadruped.
Discussion (0). Continue with ORCID to comment.