FLA · Walking: TactileStep on the Unitree G1 · 16 Nov 2026 · all reports · summary

Walking with a sensing sole: TactileStep on the Unitree G1

The other pages use the sole to read the floor. This one asks whether it can change the walk. TactileStep trains a humanoid walking policy that feels its soles and is rewarded for landing softly and using more of the sole. I rebuilt it on the Unitree G1, starting from the pretrained Hiking in the Wild policy, and compared the two at the same walking speed. The new policy did not land softer.

Simulation only (Isaac Lab). The result is a failure, stated as one.

In short

  1. Built: TactileStep's 60-point sole per foot, its training phases and its tactile rewards, on top of the Hiking in the Wild G1 task, trained from that policy for 2,000 iterations.
  2. An early pass was an artefact of speed. At iteration 400 the new policy landed with 600 N against 680 N, but it walked at 0.61 m/s instead of 0.80. Against the pretrained policy at the same speed it lost: 600 against 534 N.
  3. Speed-matched, every checkpoint fails. The final policy lands at 782 ± 10 N against 680 ± 20 N at 0.80 m/s, and uses no more of the sole (0.944 against 0.946).
  4. The gait changed. Early checkpoints land toe-first on every step; the final one lands 23 % heel, 27 % flat, 50 % toe, against 0 / 58 / 42 for the pretrained policy.
  5. Little room to gain on flat ground. The pretrained policy already uses about 95 % of the sole.
1

What was built

TactileStep is a published method for humanoid walking with tactile soles. The policy sees a pressure picture of each sole; extra rewards ask for a low impact at landing, a wide contact area and a centre of pressure well inside the foot. Hiking in the Wild is a published G1 walking policy trained on a rough-terrain (parkour) task; its released policy was the starting point.

As in the paper

Different from the paper

2

How it was tested

3

The early pass and why it was wrong

A first run on the full Hiking terrain curriculum walked, but worse than the pretrained policy (iteration 400: 834 ± 41 against 680 ± 20 N, contact area 0.820 against 0.946, 0.67 m/s). The main run then trained on flat ground and low stairs, with a check every 30 minutes that stopped training on a pass. It passed at iteration 400.

Iteration 400 against the pretrained policyActual speed (m/s)Peak landing force (N)Stance contact areaHeel / flat / toeCheck
New policy, command 0.70.611 ± 0.003600 ± 270.951 ± 0.0010.00 / 0.00 / 1.00–
Pretrained, same command 0.70.803 ± 0.002680 ± 200.946 ± 0.0020.00 / 0.58 / 0.42pass
Pretrained, command 0.5: same speed0.610 ± 0.002534 ± 150.960 ± 0.0010.01 / 0.69 / 0.30fail

Reading the table. Against the pretrained policy at the same command, the new one looked better: 80 N lower force and slightly more contact. But it walked at 0.61 m/s instead of 0.80, and slower walking lands softer by itself. Run at 0.61 m/s, the pretrained policy lands with 534 N and 0.960 contact area, so the new policy is worse on both (+66 N, −0.009). The pass was the speed, not the sole. Every later check matched speeds.

4

Speed-matched, up to 2,000 iterations

Training resumed from iteration 400 to 2,000. Each checkpoint was compared with the pretrained policy at the same actual speed. Neither policy fell in any test (0 of 24 robots).

Checkpoint (actual speed)Pretrained command → speedPeak force, new vs pretrained (N)Contact area, new vs pretrainedHeel / flat / toe, newCheck
400 (0.61 m/s)0.5 → 0.61600 ± 27 vs 534 ± 150.951 ± 0.001 vs 0.960 ± 0.0010.00 / 0.00 / 1.00fail
800 (0.74), closest0.631 → 0.74707 ± 11 vs 647 ± 60.949 ± 0.001 vs 0.953 ± 0.0010.00 / 0.00 / 1.00fail
1200 (0.77)0.665 → 0.77826 ± 34 vs 673 ± 30.947 ± 0.001 vs 0.949 ± 0.0010.00 / 0.14 / 0.86fail
1700 (0.79)0.7 → 0.80835 ± 24 vs 680 ± 200.944 ± 0.001 vs 0.946 ± 0.0020.02 / 0.48 / 0.50fail
2000, final (0.80)0.7 → 0.80782 ± 10 vs 680 ± 200.944 ± 0.000 vs 0.946 ± 0.0020.23 / 0.27 / 0.50fail

Reading the table. At every checkpoint the new policy lands harder than the pretrained one at the same speed, by 60 to 155 N, more than twice the spread over seeds. Its contact area is equal or slightly lower (differences of 0.002 to 0.009). The closest checkpoint is 800, with the smallest force excess (+60 N). In body weights, the final policy lands at 2.391 against 2.080. With more training the policy regained the speed (0.80 m/s at iteration 2,000) and its landings moved from all toe-first to a mix with some heel strikes; the foot's pitch at touchdown went from −9.3° at iteration 400 back to −2.3°, close to the pretrained −2.4°.

5

The feet, side by side

Left: the pretrained Hiking in the Wild policy. Right: TactileStep. Both commanded 0.7 m/s on flat ground, so in these clips the two do not walk at the same actual speed.

Iteration 400: the new policy walks slower and lands toe-first on every step.
Iteration 2,000, final: back to full speed, landing with a mix of heel, flat and toe.

Pretrained policy: Hiking in the Wild (CC BY-NC 4.0), shown for research.

6

What it means and its limits

Scores here are not F1: peak landing force in newtons, stance contact-area ratio, and the share of heel, flat and toe landings, each mean ± sd over 3 seeds of 8 robots on flat ground. Pretrained policy: Hiking in the Wild (CC BY-NC 4.0), shown for research; its code, data and weights are not included. Related: porting the Hiking recipe to the Booster K1, the simulated sole up close, step 1, second simulation run.