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).

**Figure 7.1.** In-house real robot task durations on a log-second axis with internal phase structure.

7.1.1.1. Look-and-Step Rough Terrain

An early speed result was in our perceptive locomotion work [64]. 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.

**Figure 7.2.** The Nadia humanoid performing a rough-terrain traversal with the *look-and-step* behavior. The composite view shows the real robot on the terrain, the 3D footstep plan, the extracted planar regions, and the first-person planar-region segmentation view used during execution. A video is available at https://youtu.be/nBMn1lJ57TU.

System	Distance (m)	Avg. speed (m/s)	Relative speed
Fallon et al. [29]	5.5	0.023	1.0x
Look-and-step [64]	5.3	0.079	3.4x
Human baseline	-	0.7	30x

Table 7.1. Rough-terrain speed comparison used in the first speed pillar claim.

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.

**Figure 7.3.** The June 23, 2021 building exploration demo on IHMC Atlas, including the hard-coded automatic pull-door behavior used as the 106 s speed anchor in [Figure 7.1](#fig:in_house_speed_phase_timeline). A video is available at https://youtu.be/CFiFaO-ENPw.

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.

**Figure 7.4.** Nadia executing the supervised can-of-soup pick-and-place behavior on June 20, 2023. Videos are available at https://youtu.be/V8jMvhVdP8k and https://youtu.be/ZBj8zs1wzik.

Time	Action completed
0:00	Begin approach.
0:11	Approach table.
0:14	Right hand approaches can.
0:53	Pre-grasp hand pose.
0:58	Grasp can of soup.
1:00	Pull back hand with can of soup.
1:16	Step to the side.
1:20	Set down can.
1:36	Release grasp on can.
1:46	Back away from task.

Table 7.2. Step-by-step supervised execution of picking and placing can of soup.

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.

**Figure 7.5.** March 15, 2024 Nadia left push-bar traversal on a door with a closer. This continuous-walking run executed in 14 s, our fastest door traversal on record. A video is available at https://youtu.be/DrR2Ng3ft5Y.

Time	Event
0:00	Approach begins.
0:04	Push bar is unlatched.
0:07	Shoulders square with the door frame.
0:14	Traversal is complete.

Table 7.3. Event times for the March 15, 2024 Nadia left push-bar traversal shown in Figure 7.5.

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.

**Figure 7.6.** The July 3, 2024 Nadia run attempting three consecutive door traversals with continuous autonomy. The robot cleared the first two doors but fell after opening the third. A video is available at https://youtu.be/cd-lo-l7pPI.

Time	Event
0:02	Start approaching door 1 (left push bar with spring closer).
0:07	Start pushing door 1 open.
0:16	Clear door 1; start approaching door 2.
0:37	Start grasping door 2 knob (right push knob door).
0:42	Open door 2.
0:53	Clear door 2; start approaching door 3.
1:18	Start grasping door 3 lever handle (right pull lever handle door).
1:21	Open door 3.
1:30	Fall during door 3 traversal.

Table 7.4. Event times for the July 3, 2024 Nadia three-door run shown in Figure 7.6.

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.

**Figure 7.7.** July 19, 2024 Nadia right pull-handle traversal with hook hands. This representative pull-side run completed in 27 s from approach start. A video of the full mock-building demo run is available at https://youtu.be/D0gylAJEdZw.

Time	Event
0:00	Approach begins.
0:07	Door is opened.
0:22	Shoulders square with the door frame.
0:27	Traversal is complete.

Table 7.5. Event times for the July 19, 2024 Nadia hook-hands right pull-handle traversal shown in Figure 7.7.

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.

**Figure 7.8.** July 19, 2024 Nadia ONR mock-building demo run 2 collage. The run entered through a push-bar door, searched multiple rooms, cleared a blocked doorway, traversed additional doors, moved furniture to recover a hidden object, and ended with a salute behavior in 7 minutes and 45 seconds. A video is available at https://youtu.be/D0gylAJEdZw.

Time	Event
0:00	Robot starts from outside first room.
0:20	Starts pushing push-bar door.
0:30	Clears right push-bar door.
0:31	Starts room search; standing in place.
0:46	Finishes room search.
0:49	Begins approach to door 2 (right pull handle).
2:30	Starts pulling open door 2.
2:49	Clears walking through door 2.
2:51	Begins searching room 2; standing in place.
3:21	Search ended; object not found; leaving room.
3:43	Starts pushing door 2 (left push bar) open to go back through.
3:54	Cleared door 2; back in main room.
4:24	After walking to center of room; search again.
5:20	Clears recycling bin blocking door 3.
5:55	Starts pulling open door 3 (right pull door).
6:13	Clears door 3.
6:37	Searches room 3; object not found; turn around.
7:00	Starts pushing door 3 (left push bar) open to go back through.
7:12	Clears door 3.
7:31	Moves couch out of the way; finds object.
7:45	Salute behavior; end of demo.

Table 7.6. Event times for the July 19, 2024 Nadia ONR mock-building demo run 2.

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.

**Figure 7.9.** Current Alex right pull lever-handle traversal key frames from the March 9, 2026 run. From left to right and top to bottom: coarse approach begins at 0:02, fine approach begins at 0:09, opening stance is achieved at 0:14, the door is unlatched at 0:18, the door is opened and traversal begins at 0:33, the shoulders align with the frame at 0:42, and traversal completes at 0:47. A video is available at https://youtu.be/fP-9DfGFvW8.

Time	Event
0:02	Coarse approach begins.
0:09	Fine approach begins.
0:14	Opening stance is achieved.
0:18	Door unlatching is completed.
0:33	Door opening is sufficient and traversal begins.
0:42	Shoulders align with the door frame.
0:47	Traversal is complete.

Table 7.7. Timestamped events for the March 9, 2026 Alex right pull lever-handle traversal shown in Figure 7.9.

Phase	Time window	Duration	Description
Approach	0:02–0:14	12 s	Coarse approach from the distant start, followed by the fine approach into the authored opening stance.
Unlatch	0:14–0:18	4 s	Final upper-body alignment, handle engagement, and latch release.
Panel motion	0:18–0:33	15 s	Pull the panel clear while staying outside the door swing.
Traverse	0:33–0:47	14 s	Commit through the frame, bring the shoulders into alignment by 0:42, and clear the doorway.

Table 7.8. Per-phase timing breakdown for the March 9, 2026 Alex right pull lever-handle traversal.

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.

**Figure 7.10.** Current Alex left push lever-handle traversal key frames from the March 9, 2026 run. From left to right and top to bottom: coarse approach begins at 0:03, fine approach begins at 0:12, opening stance is achieved at 0:17, the lever is unlatched at 0:20, the door is open and traversal begins at 0:26, the shoulders align with the frame at 0:27, and traversal completes at 0:34. A video is available at https://youtu.be/rKOUzo2sZ70.

Time	Event
0:03	Coarse approach begins.
0:12	Fine approach begins.
0:17	Opening stance is achieved.
0:20	Door lever unlatching is completed.
0:26	Door opening is sufficient and traversal begins.
0:27	Shoulders align with the door frame.
0:34	Traversal is complete.

Table 7.9. Timestamped events for the March 9, 2026 Alex left push lever-handle traversal shown in Figure 7.10.

Phase	Time window	Duration	Description
Approach	0:03–0:17	14 s	Coarse approach from the distant start, followed by the fine approach into the authored opening stance.
Unlatch	0:17–0:20	3 s	Final upper-body alignment, lever engagement, and latch release.
Panel motion	0:20–0:26	6 s	Push the panel clear while continuing to bias the body toward forward progress.
Traverse	0:26–0:34	8 s	Commit through the frame, align the shoulders with the opening by 0:27, and clear the doorway.

Table 7.10. Per-phase timing breakdown for the March 9, 2026 Alex left push lever-handle traversal.

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.

**Figure 7.11.** Alex reactive single-table ball sorting on April 4, 2026. The run includes human disturbances; the tree abandons a stale pick and returns to search when the hand-centered containment check shows that the object is no longer in the hand. A video is available at https://youtu.be/9DhgbjIRynM.

Time from start (s)	Event
0.0	Human places ball 1 (yellow) on the table.
4.5	Robot grasps ball 1.
6.9	Robot places ball 1 in container 1.
7.2	Human places ball 2 (yellow) on the table.
11.4	Robot grasps ball 2.
13.9	Robot places ball 2 in container 1.
14.5	Human places ball 3 (yellow) on the table.
17.8	Human removes ball 3 from the scene.
20.6	Robot identifies that the pick failed.
21.9	Human places ball 4 (blue) on the table.
27.4	Robot grasps ball 4.
29.8	Human places ball 5 (yellow) on the table.
30.9	Robot places ball 4 in container 2.
32.2	Human places ball 6 (yellow) on the table.
35.8	Robot grasps ball 5.
38.2	Robot places ball 5 in container 1.
42.8	Robot grasps ball 6.
45.2	Robot places ball 6 in container 1.

Table 7.11. Timeline of the Alex reactive ball-sorting demonstration. Times are relative to the first human ball placement, which occurs at 1.2 s in the source video.

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.

**Figure 7.12.** The April 14, 2026 Alex two-table loco-manipulation ball-sorting task. A video is available at https://youtu.be/KxIihoDiMtc.

Time	Event
0:00	Run begins; walk to table A.
0:10	Begin ball sorting at table A.
0:15	Grasp ball.
0:17	Place ball in container.
0:22	Grasp ball.
0:24	Place ball in container.
0:29	Grasp ball.
0:38	Walk to table B.
0:46	Begin ball sorting at table B.
0:49	Place ball carried from table A.
1:05	Grasp ball.
1:08	Place ball in container.
1:12	Grasp ball.
1:15	Place ball in container.
1:21	Grasp ball.
1:31	Walk to table A.
1:39	Begin ball sorting at table A.
1:43	Place ball in container.
1:49	Grasp ball.
1:52	Place ball in container.
1:56	Grasp ball.
2:00	Place ball in container.
2:04	Grasp ball.
2:06	Place ball in container.
2:08	Run complete.

Table 7.12. Event times for the April 14, 2026 Alex two-table loco-manipulation sorting run shown in Figure 7.12.

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.

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.

**Figure 7.14.** In-house reliability and resilience evidence at a glance. The repeated run band encodes each successful trial as a green cell (11/11, 12/12, and 32/32). The disturbance resilience band shows elapsed time phases with disturbances, retries, and opening success marked at measured timestamps (Nadia reactive pull 44 s, Alex reactive pull 65 s, ball sort 45.2 s).

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.

**Figure 7.15.** April 14, 2026 Alex left push door approach and opening reliability test. Key frames from the behavior repeated in the test. A video is available at https://youtu.be/N4rzKOKUjIA.

Run	Outcome	Phase completed
1	Success	Approach and opening
2	Success	Approach and opening
3	Success	Approach and opening
4	Success	Approach and opening
5	Success	Approach and opening
6	Success	Approach and opening
7	Success	Approach and opening
8	Success	Approach and opening
9	Success	Approach and opening
10	Success	Approach and opening
11	Success	Approach and opening

Table 7.13. Trial outcomes for the April 14, 2026 Alex left push approach and opening reliability test shown in Figure 7.15.

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 push repeat 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.

**Figure 7.16.** April 14, 2026 Alex right pull door approach and opening reliability test. Key frames from the behavior repeated in the test. A video is available at https://youtu.be/l2piKf40ea4.

Run	Outcome	Phase completed
1	Success	Approach and opening
2	Success	Approach and opening
3	Success	Approach and opening
4	Success	Approach and opening
5	Success	Approach and opening
6	Success	Approach and opening
7	Success	Approach and opening
8	Success	Approach and opening
9	Success	Approach and opening
10	Success	Approach and opening
11	Success	Approach and opening
12	Success	Approach and opening

Table 7.14. Trial outcomes for the April 14, 2026 Alex right pull approach and opening reliability test shown in Figure 7.16.

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.

**Figure 7.17.** January 2, 2026 Unitree H1-2 right pull opening reliability test. The left three panels show operator UI key frames. The right panel shows a third person view. The behavior opened the door 32 times in a row from a standing start. The test took about five minutes. A video is available at https://youtu.be/fQCNuXEF9xg.

Item	Description
Start condition	Standing at the door with the opening loop behavior loaded
Same door	One lab right pull lever door for all 32 trials
Operator	Present at the UI throughout the session
Test scope	Opening only; no approach or doorway traversal
Stop rule	Stopped after 32 successes; not run to failure
Loop structure	Goto action returns to the start of the opening sequence after each success
Measured result	32/32 successful openings
Elapsed time	About five minutes

Table 7.15. Protocol summary for the January 2, 2026 Unitree H1-2 standing right pull opening repeat test shown in Figure 7.17.

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.

**Figure 7.18.** April 12, 2024 Nadia reactive left pull handle opening under human disturbance. The robot retried five times and completed the full door traversal. A video is available at https://youtu.be/j_rzh5cAP2E.

Time	Event
0:00	Robot begins approach to left pull door.
0:06	Robot begins to grasp and turn door handle.
0:08	Human holds door closed; robot gripper slips off handle (disturbance 1).
0:11	Robot attempts 2nd grasp and turn of handle.
0:13	Human allows partial opening, then pulls door back; robot hand slips off again (disturbance 2).
0:16	With door closed again, robot attempts 3rd grasp and turn.
0:18	Human allows partial opening, pulls back; grasp slips off again (disturbance 3).
0:20	Human pushes robot arm back during 4th reach attempt, preventing grasp (disturbance 4).
0:22	Robot arm pushed during reach for 5th grasp attempt in the same way (disturbance 5).
0:26	Robot begins 6th grasp attempt.
0:29	Robot is allowed to successfully open the door.
0:34	Robot completes opening the door; begins traversal steps.
0:39	Robot shoulders even with door frame.
0:44	Robot fully clears doorway.

Table 7.16. Event times for the April 12, 2024 Nadia reactive left pull handle opening shown in Figure 7.18.

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.

**Figure 7.19.** March 13, 2026 Alex right pull door reactivity demonstration. The opening retries and blocked doorway wait are authored in the behavior tree. A video is available at https://youtu.be/pWxptD5j8b4.

Time	Event
0:01	Robot begins coarse door approach.
0:10	Robot begins fine door approach.
0:18	Robot begins 1st grasp and turn lever handle attempt.
0:19	Door partially opens; human pulls door back shut; robot grasp slips (disturbance 1).
0:22	Robot begins 2nd grasp attempt; human holds door shut; robot grasp slips (disturbance 2).
0:26	Robot begins 3rd grasp attempt; human allows partial opening again; robot grasp slips (disturbance 3).
0:30	Robot begins 4th grasp attempt; human pulls door back again; robot grasp slips (disturbance 4).
0:34	Robot begins 5th grasp attempt.
0:36	Robot is allowed to open the door.
0:45	Robot completes opening door all the way; human is now blocking doorway.
0:51	Human moves out of doorway.
0:53	Robot begins door traversal steps.
1:01	Robot shoulders even with door frame.
1:05	Robot fully clears doorway.

Table 7.17. Event times for the March 13, 2026 Alex reactive right pull lever door run shown in Figure 7.19.

7.1.2.18. Reactive Single Table Ball Sorting Disturbance

On April 4, 2026, a human removed a ball from the table while Alex was reaching to grasp it. The behavior detected the failed pick and returned to search. The robot still sorted all six balls in the run. The run is shown in Figure 7.11. Timestamps are presented in Table 7.11.

References cited on this page

[29] 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.

[64] 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.

Keyboard shortcuts

A System for Fast, Resilient, and Adaptable Loco-Manipulation Behaviors on Humanoid Robots