7.1. Hypothesis 1: Fast Loco-Manipulation Behaviors

To support our first hypothesis, we’ll present the demonstrations that highlight the system’s capabilities and breadth while also executing with high speed relative to prior and comparative works. This section also includes demonstrations of repeated-run reliability and resilience to external disturbances when trying to traverse doors and pick and place objects.

7.1.1. Speed and Capability Evidence

In Figure 7.1, we present a representative sample of our loco-manipulation behaviors. This figure shows that our door behaviors were very slow on Atlas, over a minute, when we had hard-coded behaviors. Door behaviors were drastically sped up on Nadia with runtime-editable structure and concurrent action layering. Later on Alex, we demonstrate new manipulation tasks and locomotion transitions in the same speed regime of seconds to minutes.

In the figure, the rows are color coded to show different phases of the behaviors. For door behaviors, it shows approach, opening, and traversal phases, and for loco-manipulation behaviors it shows walking and manipulation phases. Triangle and square markers indicate object grasps and placements. The dashed vertical line marks the fastest IHMC door traversal on record (14 s, Nadia 2024 left push-bar).

7.1.1.1. Look-and-Step Rough Terrain

An early speed result was in our perceptive locomotion work [71]. Figure 7.2 illustrates a 14-step key result, taking 14 autonomous steps in 37 seconds. The speed of this behavior compared to prior work in the literature and an estimated human baseline is presented in Table 7.1.

This 14 step run averaged 0.079 m/s over 5.3 m (Table 7.1), about 3.4 times faster than the Fallon et al. baseline at 0.023 m/s. This demonstration uses an earlier architecture than is tested in hypothesis 1. It is still useful context because it shows an early focus as part of this thesis in achieving faster robot locomotion and onboard sensing.

7.1.1.2. Atlas Hard-Coded Pull-Door Traversal

Our earliest in-house door-speed anchor is the hard-coded IHMC Atlas pull-door traversal from the June 2021 building exploration era, shown in Figure 7.3. That step-by-step behavior took 106 s from approach start to full traversal completion, with long pauses, fiducial dependence, and no robot-local authored recovery.

Compared with the July 2024 Nadia right pull handle traversal at 27 s (Table 7.5), this Atlas run is about 3.9 times slower on a pull side door task. The later run removed fiducial dependence, used runtime authored structure, and was run on a different platform. The comparison is not controlled, but still anchors how much faster our recent door behaviors became relative to this IHMC baseline.

7.1.1.3. Supervised Can Pick-and-Place

On June 20, 2023, we executed a supervised can-of-soup pick-and-place behavior on Nadia in 1 minute and 46 seconds, shown in Figure 7.4. The run was operator-supervised, used an ArUco marker rather than direct can detection, and required manual gripper retries between actions. Timestamps are presented in Table 7.2.

The 1 minute 46 second run shows an early manipulation capability on Nadia, but much of the timeline is operator supervised tuning rather than autonomous execution (Table 7.2). ArUco marker dependence and operator supervision make this demonstration an early speed reference rather than a testable result for hypothesis 1. It does show that the architecture could, albeit with great difficulty, execute pick and place on real hardware before YOLO based scene authoring was in place.

7.1.1.4. 14 s Nadia Left Push-Bar Traversal

Our fastest door traversal ever was on March 15, 2024, and is presented in Figure 7.5. The robot walked continuously during this run, using concurrent arm actions to open the push bar door during the traversal. Timestamps are presented in Table 7.3.

The robot unlatched the push bar at 4 s and cleared the doorway at 14 s (Table 7.3). Continuous walking with concurrent arm actions produced our fastest door traversal on record and is marked by the dashed line in Figure 7.1. Relative to the 106 s Atlas anchor, this run is direct evidence that concurrent action layering removed long execution pauses.

7.1.1.5. Three Doors in a Row

On July 3, 2024, we attempted to traverse three lab doors consecutively in one continuous run, shown in Figure 7.6. The robot cleared the first two doors but fell during traversal of the third; the timed run ended at 1 minute and 30 seconds. Timestamps are presented in Table 7.4.

The robot cleared the first door at 16 s and the second at 53 s, then fell during the third traversal at 1 minute 30 s (Table 7.4). The run still supports capability breadth across push bar, push knob, and pull handle doors in one continuous behavior. The fall shows that full traversal reliability remained a weak point, which we talk about in the reliability results below.

7.1.1.6. Nadia Right Pull-Handle Traversal

On July 19, 2024, Nadia completed a representative right pull-handle door traversal with hook hands in 27 seconds, shown in Figure 7.7. Timestamps are presented in Table 7.5.

The 27 s total matches the pull lever traversal measurement in Figure 7.1. Door opening finished at 7 s, but panel clearance kept the robot from getting through the frame until 22 s (Table 7.5). Figure 7.13 shows the same push versus pull timing pattern across Nadia runs.

7.1.1.7. ONR Mock-Building Demo Run 2

On July 19, 2024, we executed the second timed full run of our ONR mock-building search demo in 7 minutes and 45 seconds, shown in Figure 7.8. This run searched three rooms, cleared a blocked doorway, moved furniture, and ended with a salute behavior. Timestamps are presented in Table 7.6.

The 7 minute 45 second run combined three room searches, five door passages, furniture manipulation, and a blocked doorway recovery (Table 7.6). Much of the elapsed time went to searching and repositioning between doors, not to isolated door traversals. The standalone 27 s pull handle run from the same day (Table 7.5) shows that individual door phases could be fast, but this chained demo mainly demonstrates breadth across many authored behaviors in one tree.

7.1.1.8. Alex Right Pull Lever Traversal

On March 9, 2026, Alex completed a right pull lever-handle door traversal in 45 seconds, shown in Figure 7.9. Timestamps are presented in Table 7.7.

This 45 s run breaks down into 12 s approach, 4 s unlatch, 15 s panel motion, and 14 s traverse (Table 7.8). Panel motion is the longest phase on pull doors because the robot must stay clear of the swing. This timing places Alex pull traversals in the same tens of seconds band as our 2024 Nadia results and well below the minute scale classical literature entries in Figure 7.29.

7.1.1.9. Alex Left Push Lever Traversal

On March 9, 2026, Alex completed a left push lever-handle door traversal in 34 seconds, shown in Figure 7.10. Timestamps are presented in Table 7.9.

The 34 s push side run is faster than the 45 s pull run on the same robot and week. Approach and panel motion together take only 20 s, and shoulders align with the frame just 1 s after traversal begins (Table 7.10). This matches the progress curve pattern in Figure 7.13, where push doors allow more continuous forward motion.

7.1.1.10. Reactive Single-Table Ball Sorting

On April 4, 2026, Alex completed a reactive single-table ball-sorting run in 45.2 seconds while humans continuously placed and sometimes removed balls on the table, shown in Figure 7.11. Timestamps are presented in Table 7.11.

The 45.2 s run sorted six balls while humans continuously placed new balls on the table (Table 7.11). This rate of action supports hypothesis 1 on manipulation speed under a dynamic scene. Resilience to the ball removal disturbance at 17.8 s is analyzed in 1.2.6.

7.1.1.11. Two-Table Loco-Manipulation Sorting

On April 14, 2026, we extended the reactive ball-sorting behavior to a two-table loco-manipulation task, shown in Figure 7.12. In the timed execution run used in Figure 7.1, the robot sorted nine balls across two tables in 2 minutes and 8 seconds. Timestamps are presented in Table 7.12.

Nine balls were sorted between two tables in 2 minutes 8 seconds, with locomotion between tables at 0:00, 0:38, and 1:31 (Table 7.12). This demonstration shows a loco-manipulation capability beyond door traversal and exploration tasks. The timed run appears in Figure 7.1; the behavior extension that produced it is analyzed under hypothesis 3.

7.1.1.12. Nadia Door-Traversal Progress Curves

Figure 7.13 illustrates representative forward progress curves for door traversals on Nadia, showing the difference between push and pull doors. Pull doors can take longer because the robot must stay clear of the panel as it opens.

The four Nadia curves separate push doors, which keep moving forward, from pull doors, which stall progress while the panel opens. The 14 s push bar and 27 s pull handle runs above match these two curve shapes. The Alex push and pull traversals above follow the same qualitative pattern at 34 s and 45 s, even though they are not plotted in this figure.

Figure 7.1 summarizes the speed arc from the 106 s Atlas anchor through 14 to 45 s door traversals and minute scale manipulation tasks. The individual runs above supply the timed events behind each bar. Together they support the speed and capability parts of hypothesis 1.

7.1.2. Reliability and Resilience Evidence

In Figure 7.14, we summarize our main in-house reliability and resilience anchors. The top band shows consecutive successful trials for repeated door approach and opening on Alex and Unitree H1-2. The bottom band shows disturbed runs where the robot retried and recovered. Photos and event tables for each run follow in the subsections below.

7.1.2.13. Alex Left Push Approach and Opening Repeat

On April 14, 2026, we ran the left push door behavior on Alex with a goal of at least 10 approach and opening trials. We lost count during the session and kept going so we would not stop short. The test ended with 11 successes in a row. All trials used the same lab left push door on the same night. An operator was present at the UI. We did not run the test to failure. Each run completed the approach and opening phases without failure. We did not repeat full doorway traversal in this test. The test is shown in Figure 7.15. Trial outcomes are presented in Table 7.13.

Eleven consecutive approach and opening trials completed without failure (Table 7.13). We did not test full traversal because the walking controller was not working well for traversal at the time of these experiments. However, our repeated trials still demonstrate reliable approach and opening. This 11/11 result is the Alex left push row in Figure 7.14.

7.1.2.14. Alex Right Pull Approach and Opening Repeat

On April 14, 2026, we ran the right pull door behavior on Alex with a goal of at least 10 approach and opening trials. We lost count during the session and kept going so we would not stop short. The test ended with 12 successes in a row. All trials used the same lab right pull door on the same night as the repeated push test. An operator was present at the UI. We did not run the test to failure. Each run completed the approach and opening phases without failure. We did not repeat full doorway traversal in this test. The test is shown in Figure 7.16. Trial outcomes are presented in Table 7.14.

Twelve consecutive approach and opening trials completed without failure on the same night as the repeated push test (Table 7.14). Again, the scope was approach and opening only. This 12/12 result is the Alex right pull row in Figure 7.14.

7.1.2.15. Unitree H1-2 Standing Right Pull Opening Repeat

On January 2, 2026, we ran the standing right pull lever opening behavior 32 times in a row on Unitree H1-2. All trials used the same lab right pull door during the same session. An operator was present at the UI. We stopped after 32 successes; we did not run the test to failure. The robot opened the door from a standing start each time. The full test took about five minutes. The test is shown in Figure 7.17. The test protocol is presented in Table 7.15.

The robot opened the same lab door 32 times in a row from a standing start in about five minutes (Table 7.15). This repeated trial was autonomous even between repetitions: the goto loop returned to the same opening sequence after each success. It also serves as evidence that our system’s reliability extends beyond a single humanoid robot platform. The 32/32 result is the Unitree row in Figure 7.14.

7.1.2.16. Nadia Reactive Left Pull Handle Opening

On April 12, 2024, we disturbed a Nadia left pull door opening five times during one run. The hard coded door traversal node detected each failure and retried until the door opened. The robot finished the full traversal without operator intervention. The run is shown in Figure 7.18. Timestamps are presented in Table 7.16.

Five human disturbances during opening caused grasp slips or blocked reaches (Table 7.16). A hard coded traversal node retried until the door opened at 29 s and the robot cleared the doorway at 44 s without operator intervention. This is an earlier resilience mechanism than the authored fallback trees used in later Alex behaviors, but it still produced the Nadia reactive pull row in Figure 7.14.

7.1.2.17. Alex Reactive Right Pull Lever Door

On March 13, 2026, we disturbed an Alex right pull lever door opening four times during the opening phase. A human also blocked the doorway before traversal. The behavior retried the opening and waited until the doorway was clear. The run is shown in Figure 7.19. Timestamps are presented in Table 7.17.

Four disturbances during opening and a blocked doorway before traversal triggered retries and a wait for clearance (Table 7.17). The run finished in 65 s overall. Unlike the 2024 Nadia case, the retry and wait logic here was authored in the behavior tree with fallback and condition nodes, and it appears as the Alex reactive pull row in Figure 7.14.

7.1.2.18. Reactive Single Table Ball Sorting Disturbance

For the previously presented 45.2 s sorting run in 1.1.10, the ball removal at 17.8 s is the main resilience event. The reactive tree logic abandoned the stale pick at 20.6 s and still placed all six balls by 45.2 s. That recovery is the ball sorting row in Figure 7.14.

The repeated run and disturbance results above support reliability and resilience within hypothesis 1. We do not have comparable repeat trial data for prior IHMC door baselines. Our tests also stopped short of repeated full door traversals because walk through performance was not yet trustworthy enough to risk the hardware.

References cited on this page

[33] M. F. Fallon et al., “Continuous humanoid locomotion over uneven terrain using stereo fusion,” in 2015 IEEE-RAS 15th international conference on humanoid robots (humanoids), 2015, pp. 881–888.

[71] D. Calvert, B. Mishra, S. McCrory, S. Bertrand, R. Griffin, and J. Pratt, “A fast, autonomous, bipedal walking behavior over rapid regions,” in 2022 IEEE-RAS 21st international conference on humanoid robots (humanoids), 2022, pp. 24–31. doi: 10.1109/Humanoids53995.2022.10000120.