Figure Shows Figure 03 in Autonomous Ladder-Climb Demo

The video shows Figure 03 approaching a ladder and ascending rung by rung. That is evidence of one ladder-climb sequence under the demonstrated conditions. The autonomy description and operating constraints remain company claims, not results from an independently observed or reproducible evaluation.

Ladder climbing is a demanding mobility task because the robot must coordinate hand and foot placement across narrow, intermittent contacts while controlling its center of mass. Compared with stair ascent, a ladder also gives the system less support area and less room to recover from a missed rung or positioning error.

Figure's separate April 29 production update provides relevant background rather than evidence specific to this ladder task. The company said its BotQ facility had delivered more than 350 third-generation robots and increased Figure 03 production from one robot per day to one per hour. It also reported an end-of-line first-pass yield above 80% and more than 150 networked workstations supporting production.

In the April update, Figure said Helix System 0 uses head-camera RGB input to build a 3D representation of the robot's surroundings and conditions whole-body control on that representation. The company said the policy was trained with reinforcement learning on randomized terrain in simulation and transferred to real-world stair traversal without real-world fine-tuning or operator mode changes.

That description does not establish that the ladder task used the same training setup, policy, sensors, or evaluation conditions. Figure has not publicly disclosed success rates, number of trials, ladder variations, safety mechanisms, or whether the robot had encountered the configuration before.

The clip is a technically relevant mobility demonstration, but it does not establish robust autonomous vertical access in industrial environments. A stronger evaluation would vary rung spacing, ladder angle, lighting, surface conditions, payload, and approach geometry while reporting repeatability and recovery after missed contacts.

Useful evidence would also include contact-force telemetry, perception latency, fall-prevention behavior, human intervention rules, and results across multiple robots. Until those details are available, the narrow conclusion is that Figure demonstrated one company-described autonomous ladder ascent, not that Figure 03 has a generally validated ladder-climbing capability.