Behaviour commands: the swing-height command works

The swing-height command is followed once its reward term is scaled to Pupper's step height and the policy is trained with trot commands only. The same policy then shows what a higher step does on obstacles.

• Swing height at an 8 cm command, runs of 200 M steps: 4.9 cm (first setup), 5.4 cm (weight × 10), 5.9 cm (trot commands only), 7.1 cm (trot commands only, weight × 40); the line set before the runs was 6.5 cm
• Why the weight alone did nothing: under a pace command the policy keeps trotting, and three quarters of the swing term then comes from feet on the ground while the command wants them in the air
• A higher step alone, same policy at an 8 cm against a 5 cm command, 100 trials: 7 cm curb 22 → 43 passes, 3 cm stairs on 80 mm treads 94 → 75, 4 cm boxes 4 → 10
• Pace, bound and pronk are not followed in any run; whether they stay in the command set is open