Learning contactfrom motor current

Most people learn a manual skill by watching someone else, trying it, or by guidance from an expert. A robot has traditionally required an engineer to describe much of that behaviour in advance. In imitation learning, a person demonstrates a task and a policy learns the relationship between what it observes in the environment and how it should act. Teaching a robot a new manipulation skill then becomes similar to teaching by showing.

The promise of learning from demonstrations is easiest to realise when the task can be inferred from geometry. In free space, a camera shows the location of objects and joint positions show the position of the arm in space. Contact-rich manipulation is harder because the decisive state may be physical and no longer purely geometric. A gripper closes without securing an object. An object slips midway. A peg reaches the hole and presses against its rim. Two nearly identical camera images precede a smooth insertion in one case and a jam in the other.

Low-cost bus servos already expose a motor-current register. The Waveshare ST3215 control table specifies a read-only motor-current register. Reading an existing current register requires no additional sensing hardware.

Visual occlusion

This loss of direct visual evidence near contact is a central observability problem in contact-rich imitation learning.

1. 1

   Images and joint positions provide strong information about appearance, configuration, and motion. They do not directly measure interaction effort.

2. 2

   Contact-aware policies add tactile, force, or torque measurements to vision and kinematics. These measurements reveal contact onset, load, deformation, or slip, and they usually require added hardware, calibration, or a robot that already gives reliable torque estimates.

3. 3

   The platform is an SO-101 arm fitted with Waveshare ST3215 bus servos. Their control table specifies a read-only motor-current register. Reading that existing register requires no additional sensing hardware.

4. 4

   Motor-side current is affected by pose, motion, friction, and the high-ratio transmission. It does not specify the direction or location of an external force.

Three questions follow from the overarching one. They separate three claims that are easy to confuse. The signal may contain task-relevant information. A trained model may use that information. Using it may improve closed-loop behaviour.

Characterisation of the current signal

Before motor current can be added to a policy, it must first be shown to carry a repeatable and task-relevant signal.

Dataset

A fair comparison between a vision-based policy and a current-aware policy requires more than recording both modalities. The policies must learn from the same demonstrations, images, joint positions, and action targets, while the additional current input is introduced through a reproducible transformation.

Evaluation

Binary task success is the primary endpoint. A trial succeeds when the peg is fully seated and flush with the fixture surface before the timeout. Every other outcome is a failure. One operator resets and scores every trial.

Results

The strongest favourable task result is obtained with joint-selective temporal ACT. It succeeds in 12/30 trials against 7/30 for its contemporaneous vision-and-position control, a 71.4% relative improvement and a gain of 16.7 percentage points. The Diffusion Policy comparison also favours the current-aware design, with 12/30 successes against 9/30, a 33.3% relative improvement. Direct current concatenation produces lower observed success in both the raw and normalised ACT comparisons, with relative decreases of 57.9% and 11.8%.

The favourable temporal ACT result is of practically relevant observed size, although its difference interval includes zero and does not establish superiority. The current-aware policy scores higher in three blocks and ties its control in two. These blocks describe the same checkpoint pair within one logical evaluation session, completed over two dates. They are not independent replications. Since visual-baseline success varies across sessions, the within-session relative gains do not establish a controlled ranking of the current representations. In the Diffusion Policy session, combined success falls from 8/12 in the first block to 6/12, 4/12, 2/12, and 1/12. The shared within-session deterioration is stronger than the observed between-policy difference and limits interpretation of either absolute rate.

Conclusion

Existing motor-current readings on the evaluated SO-101 contain contact-related information and can influence learned action predictions. The strongest favourable task result comes from joint-selective temporal ACT: success increases from 23.3% to 40.0%, a 71.4% relative improvement over its contemporaneous vision-and-position control. This observation supports further investigation of the current-aware setting, while its uncertainty leaves superiority unresolved.

Citation

@mastersthesis{ranpariya2026contact,
  author = {Jaydeepkumar Ranpariya},
  title  = {Learning Contact Representations from Zero-Cost
            Proprioceptive Effort Signals for Contact-Rich
            Visuomotor Imitation Learning},
  school = {Friedrich-Alexander-Universit{\"a}t Erlangen-N{\"u}rnberg},
  year   = {2026},
  note   = {In cooperation with Fraunhofer IIS},
  url    = {https://contact-from-current.pages.dev/}
}