What's the real-world accuracy of visual odometry on a walking humanoid?

[niklassantos]

@scott.andersson5 This matches something I went through recently. A minimum viable sensing suite for safe bipedal walking generally includes joint encoders, an IMU for orientation/angular velocity, and either force/torque sensing or accurate current-based torque estimation at the ankles - everything else (vision, tactile, LiDAR) adds capability rather than being strictly required just to stay upright. IMU drift over time (bias instability) is usually the real culprit behind slowly diverging state estimates, not noise - it's typically handled with sensor fusion against other references (visual odometry, joint kinematics) rather than trying to eliminate drift at the source.

[deborah59]

This lines up with my experience. Multi-camera calibration drifts over time from thermal expansion, vibration, and mechanical wear, which is why production systems typically run periodic recalibration routines rather than assuming a one-time factory calibration holds forever.

[timothy.roberts2]

I don't think that's quite right, for what it's worth. Estimating joint torque from motor current draw is cheap and requires no extra sensor, but it's less accurate than a dedicated torque sensor because it doesn't capture friction losses through the gearbox - good enough for coarse control, not always for precise force-controlled tasks. Reminds me a bit of the early drone hobbyist scene, honestly.

[chloe_jack]

Here's the relevant bit as far as I understand it: A minimum viable sensing suite for safe bipedal walking generally includes joint encoders, an IMU for orientation/angular velocity, and either force/torque sensing or accurate current-based torque estimation at the ankles - everything else (vision, tactile, LiDAR) adds capability rather than being strictly required just to stay upright. Multi-camera calibration drifts over time from thermal expansion, vibration, and mechanical wear, which is why production systems typically run periodic recalibration routines rather than assuming a one-time factory calibration holds forever.

[richard36]

Genuine beginner question - Latency between a perceived event (like a slip) and a corrective control response matters enormously for balance - even 50-100ms of extra perception latency can be the difference between a smooth recovery and a fall, which is part of why a lot of balance-critical sensing is proprioceptive rather than vision-based. Makes me wonder how this looks in another five years.

[emma_whit]

+1 to this. Worth adding: Proprioception (the robot's sense of its own joint angles, velocities, and forces) tends to get less attention than flashy vision systems, even though a lot of balance and manipulation failures trace back to proprioceptive noise or miscalibration rather than a vision problem. Vibration is one of the most underrated sources of noisy IMU and tactile readings - mounting matters as much as sensor quality, and a poorly isolated mount can add more noise than the sensor's own datasheet specs would suggest.

[timothy.roberts2]

@emma_whit Not sure I fully agree here. Proprioception (the robot's sense of its own joint angles, velocities, and forces) tends to get less attention than flashy vision systems, even though a lot of balance and manipulation failures trace back to proprioceptive noise or miscalibration rather than a vision problem. Kind of makes me think about how different this all looked even three years ago.

[smartinez]

@timothy.roberts2 This matches something I went through recently. IMU drift over time (bias instability) is usually the real culprit behind slowly diverging state estimates, not noise - it's typically handled with sensor fusion against other references (visual odometry, joint kinematics) rather than trying to eliminate drift at the source. Event cameras (which report per-pixel brightness changes rather than full frames) are still more of a research curiosity than a production sensor for humanoids, mainly because the software ecosystem and processing pipelines around them are far less mature than for standard frame-based cameras.

[matthew43]

@smartinez From what I've seen: LiDAR gives reliable, lighting-independent range data but is heavier, pricier, and gives sparser point clouds up close than stereo or depth cameras, which is why a lot of humanoids lean on stereo/depth cameras for near-field manipulation and reserve LiDAR (if present at all) for longer-range navigation. Multi-camera calibration drifts over time from thermal expansion, vibration, and mechanical wear, which is why production systems typically run periodic recalibration routines rather than assuming a one-time factory calibration holds forever.

[wei_ross]

@matthew43 Same conclusion I've come to. Also worth noting: Event cameras (which report per-pixel brightness changes rather than full frames) are still more of a research curiosity than a production sensor for humanoids, mainly because the software ecosystem and processing pipelines around them are far less mature than for standard frame-based cameras. IMU drift over time (bias instability) is usually the real culprit behind slowly diverging state estimates, not noise - it's typically handled with sensor fusion against other references (visual odometry, joint kinematics) rather than trying to eliminate drift at the source.