5.2.5.5 Testing & Evaluation
Performance Criteria
The following measurable targets were established from the torque calculations (Appendix 1), padwork research, and the instrumentation available within the project scope:
| ID | Criterion | Target | How Measured |
|---|---|---|---|
| ARM-AC-1 | Strike speed (90° rotation) | ≤ 0.25 s | Firmware timestamp delta (N=43 speed tests) |
| ARM-AC-2 | Sparring endurance | 5 min continuous, <60°C motors | Continuous sparring session with Damiao temperature telemetry |
| ARM-AC-3 | Regenerative braking safety | No PSU trip over 5 emergency stops | Visual confirmation and voltmeter monitoring |
| ARM-AC-4 | Multi-strike chain | 3-strike chain ≤ 5 s total | Video timestamp and GUI console log |
| ARM-AC-5 | Peak current within safety limit | < 2.0 A per motor across all strikes | CAN current telemetry (N=43 speed tests) |
Test Results
Results from the speed test validation (N=43 tests, all 6 standard strikes) are presented below. ARM-AC-2, ARM-AC-3, and ARM-AC-4 require dedicated test sessions and are documented as pending.
| ID | Criterion | Target | Measured | Met? |
|---|---|---|---|---|
| ARM-AC-1 | Strike speed (90°) | ≤ 0.25 s | 0.64 s (best, 30 rad/s Jab) | Partial |
| ↓ See Speed Test Benchmark below for full analysis of the PID overhead bottleneck. | ||||
| ARM-AC-2 | Sparring endurance | 5 min, <60°C | N/A | Pending |
| ARM-AC-3 | Regen braking safety | No PSU trip / 5 E-stops | N/A | Pending |
| ARM-AC-4 | Multi-strike chain | 3-strike ≤ 5 s | N/A | Pending |
| ARM-AC-5 | Peak current safety | < 2.0 A per motor | 0.69 A max (Right Hook M4) | Pass |
Speed Test Benchmark (N = 43 Tests)
A controlled strike-speed benchmark was conducted using the Speed Test tab across all 6 standard strikes at set speeds ranging from 10 to 30 rad/s, with 3 repetitions each. All 43 tests passed with zero safety trips (no current-limit events, no position timeouts).
pk_rad_s
and avg_rad_s are motor-shaft angular velocity.
Per-Strike Summary (at 25 rad/s, 3 reps each)
| Strike | Arm | Avg WU→AP (s) | Avg Total (s) | Pk Motor RPM | Pk Current (A) | Avg Sys Power (W) |
|---|---|---|---|---|---|---|
| Jab | Left | 0.66 | 1.24 | 239 | M1: 0.63, M2: 0.58 | ~10.5 |
| Left Hook | Left | 0.67 | 1.25 | 300 | M1: 0.57, M2: 0.59 | ~9.7 |
| Left Uppercut | Left | 0.55 | 1.09 | 218 | M1: 0.49, M2: 0.57 | ~10.1 |
| Cross | Right | 0.61 | 1.18 | 285 | M3: 0.61, M4: 0.63 | ~9.4 |
| Right Hook | Right | 0.62 | 1.20 | 207 | M3: 0.51, M4: 0.69 | ~8.6 |
| Right Uppercut | Right | 0.57 | 1.07 | 238 | M3: 0.52, M4: 0.62 | ~8.9 |
Speed vs Execution Time (Left Jab)
| Set Speed (rad/s) | Avg WU→AP (s) | Avg Pk Motor RPM | Arm Joint RPM |
|---|---|---|---|
| 10 | 0.85 | 113 | ~38 |
| 15 | 0.73 | 185 | ~62 |
| 20 | 0.68 | 223 | ~74 |
| 25 | 0.66 | 226 | ~75 |
| 30 | 0.64 | 254 | ~85 |
Diminishing returns above 20 rad/s: execution time plateaus near 0.64 s as the PID acceleration/deceleration ramp time dominates.
PID Overhead Analysis: Why the Original 0.25 s Target Was Not Achievable
Triangular Velocity Profile
A 90° arm sweep (~1.57 rad joint / ~4.71 rad motor) at constant velocity theoretically requires only 0.19 s at 25 rad/s. However, the Damiao DM-J4310-2EC internal PID controller applies acceleration and deceleration ramps that add approximately 0.4–0.5 s of constant overhead regardless of the set speed:
| Set Speed | Theoretical (const vel) | Measured WU→AP | Overhead |
|---|---|---|---|
| 10 rad/s | 0.52 s | 0.85 s | +63% |
| 20 rad/s | 0.26 s | 0.68 s | +162% |
| 30 rad/s | 0.17 s | 0.64 s | +276% |
The velocity profile is a triangle (not trapezoid): the motor never reaches steady-state velocity before deceleration begins. This is the primary bottleneck for strike speed. Tuning the Damiao PID acceleration parameters or implementing feed-forward control could reduce this overhead.
Key Findings
- Current draw is very low: peak per-motor current never exceeds 0.69 A (Right Hook M4). Well below the 2.0 A safety limit and 3.0 A firmware watchdog.
- Total arm power is approximately 33 W peak and 10 W average across 4 motors combined. Cross-arm leakage is negligible (0.01–0.10 A).
- Uppercuts are fastest: shortest travel distance yields approximately 0.55 s warm-up to apex, versus approximately 0.66 s for the Jab at the same set speed.
- No safety trips occurred in 43 tests: zero current limit events and zero position timeouts.
Power Budget Validation
The 43-test speed benchmark measured peak arm motor power at approximately 33 W, substantially below the theoretical maximum of 384 W (16 A stall current at 24 V). Three factors account for the order-of-magnitude discrepancy:
- Short travel distance (dominant): A punch is ~90° of arm sweep = ~4.71 rad at the motor. At 25 rad/s, the theoretical constant-velocity transit is only 0.19 s. The motor does not have sufficient distance to sustain peak velocity.
- PID acceleration/deceleration overhead: The approximately constant ~0.4–0.5 s ramp overhead means the velocity profile is triangular: the motor never reaches steady-state before deceleration begins.
- Current ∝ torque, not speed: The 16 A figure is the stall current (locked rotor, zero speed, maximum torque). During free punches actuating lightweight pool noodle padding (~200 g per arm), the torque demand never approaches stall values. Typical per-motor current is 0.5–1.0 A during acceleration, near-zero during constant velocity, and negative (regenerative) during deceleration.
Dual-Purpose Power Figures for Thesis
| Value | Purpose | Cite As |
|---|---|---|
| 33 W peak / 10 W avg | PSU sizing, thermal design, operating cost, efficiency analysis | “Measured operating power under sparring conditions (N=43 tests)” |
| 384 W (theoretical stall) | Wire gauge selection, fuse rating, current-limit design, safety analysis | “Theoretical maximum under locked-rotor fault condition” |
ros2_ws/unified_v4/strike_speed_results.csv.
Video Evidence
Limitations
- Strike speed target revised: The original 0.25 s target assumed constant-velocity traversal. The Damiao PID controller’s acceleration/deceleration ramp adds ~0.4–0.5 s of irreducible overhead. The system target of ≤ 0.25 s remains as a formal requirement (RM-8); the best recorded result of 0.64 s constitutes a partial fulfilment. Feed-forward control could further reduce this overhead.