Learning from human demonstrations has enabled robots to acquire a wide range of manipulation skills, but learned policies typically execute far slower than ordinary humans. This speed gap is mainly due to lack of an interface for collecting demonstration data at high speed, and the difficulty in training policies that can robustly execute high-speed motions. In this paper, we present ALOHA Lightning (Flash ALOHA), a system for learning fast and precise robotic manipulation. Our system uses kinesthetic teaching to intuitively collect near-human-speed demonstrations on a backdrivable bimanual platform, yielding natural and fast trajectories. We also present a learning pipeline that enables smooth high-speed execution through test-time action smoothing and aligns the visual data distribution between data collection and deployment with masking. Given 50 demonstrations for each task, ALOHA Lightning autonomously completes bimanual tasks such as ball tossing and catching, folding shorts, battery insertion, and bussing tables for over 80% success rates while close to human speed. Project website: https://flash-aloha.github.io/
Figures & tables
Fig. 2 : Hardware Setup. Our printed handles allow for easy operation and data collection.
Fig. 3 : Masking. Our masking method adjusts in real time based on the position of the end effectors of both arms.
Fig. 4 : Test-Time Smoothing. Given a previous chunk of target joint positions q^1 and a later chunk q^2 , we apply smoothing to obtain the final commanded actions a .
Fig. 5 : Masking Comparisons. We show the same initial starting frame of Battery Insertion . Our masking method completely aligns the visual distributions of training and test times. Inpainting via diffusion creates unwanted artifacts.
Masking
Unstack Cups
Battery Insertion
Bus Table
Fold Shorts
Method
Grab
Pull
Whole
Time(s)
Pick
Place
Prepare
Push
Whole
Time(s)
C&S
B&P
Whole
Time(s)
Up
Right
Whole
Time(s)
No Mask
90
90
80
4.12
95
95
85
75
75
6.44
100
100
100
6.97
100
90
90
6.19
Inpainting
100
85
85
3.92
95
95
70
65
65
5.65
100
100
100
6.57
100
80
80
6.20
Ours
100
90
90
3.97
100
100
90
85
85
5.70
100
100
100
6.81
100
100
100
5.94
TABLE I : Our masking improves performance. Our masking strategy almost uniformly beats non-masked and diffusion inpainted in policies, enabling tasks such as battery insertion.
Smoothing
Unstack Cups
Battery Insertion
Bus Table
Fold Shorts
Method
Grab
Pull
Whole
Time(s)
Pick
Place
Prepare
Push
Whole
Time(s)
C&S
B&P
Whole
Time(s)
Up
Right
Whole
Time(s)
None
100
65
65
4.23
100
85
75
75
75
5.29
100
80
80
6.70
95
80
80
6.16
TE
95
35
35
7.75
100
100
85
55
55
10.18
100
100
100
12.23
100
90
90
11.60
Ours
100
90
90
3.97
100
100
90
85
85
5.70
100
100
100
6.81
100
100
100
5.94
TABLE II : Smoothing improves performance. Test time smoothing outperforms or matches other methods in success rate in every category
Speed Comparisons
Unstack Cups
Battery Insertion
Bus Table
Toss & Catch
Method
Grab
Pull
Whole
Time(s)
Pick
Place
Prep
Push
Whole
Time(s)
C&S
B&P
Whole
Time(s)
Pick & Toss
Catch
Whole
Time(s)
Puppeteering
85
75
75
8.84
100
90
60
55
55
17.87
100
90
90
16.05
–
–
–
–
Puppet Double Freq
80
65
65
5.47
80
65
35
15
15
12.90
95
85
85
9.30
–
–
–
–
Puppet Skip Actions
85
50
50
5.58
95
95
55
30
30
10.30
95
85
85
9.29
–
–
–
–
Ours
100
90
90
3.97
100
100
90
85
85
5.70
100
100
100
6.81
100
50
50
1.85
TABLE III : Speed Comparisons. We note a sharp drop in performance on precise tasks when applying either speed-up approach.
Fig. 6 : User Study. Results on Battery Insertion for timing operators on both Ours and Puppeteering demonstrate a 50%+ gap in duration for both experienced and new operators.
Robot manipulation policies trained through imitation learning inherit not only the demonstrated behavior but also the conservative execution tempo of robot demonstrations. Existing acceleration approaches can execute faster than the original demonstrations, but determine the appropriate acceleration primarily from robot-side information or a predefined set of tempo factors, leaving open how to obtain a task-appropriate reference for how fast each manipulation phase should progress. We introduce EgoSpeedUp, a framework that uses human manipulation as temporal supervision for robot imitation learning. Our key insight is that human demonstrations naturally reveal task-appropriate, phase-wise manipulation tempo. Given slow robot demonstrations and human demonstrations of the same task, EgoSpeedUp aligns corresponding manipulation phases, estimates their relative execution tempos from multiple human demonstrations, and transfers the resulting phase-wise tempo by retiming the robot demonstrations. The retimed demonstrations are then used for standard behavior cloning, allowing the robot to retain its executable manipulation behavior while learning to perform it at a human-informed tempo. Across two real-world manipulation tasks, EgoSpeedUp improves the task success rate by an average of 25 percentage points (pp) while reducing successful execution time by 36.5%. These results demonstrate that human manipulation tempo provides an effective temporal reference for learning faster and more reliable robot policies.
Hanbit Oh, Yukiyasu Domae, Takuma Yagi
Artificial Intelligence Research Center, National Institute of Advanced Industrial Science and Technology (AIST), Japan
Learning whole-body mobile manipulation requires demonstrations that coordinate bimanual actions with locomotion across extended workspaces. However, existing interfaces trade off direct robot control, perception and action alignment, and natural operator mobility, making such demonstrations difficult to collect. We introduce HALO, a human-aligned data collection system for whole-body mobile manipulation. The system aligns the operator's perception and action space with the robot while preserving direct physical feedback and natural locomotion. Operators guide the robot end-effectors through kinesthetic teaching and move a low-resistance platform without the deployed mobile base while performing bimanual manipulation. This interface enables more natural whole-body demonstrations, but the close human-robot coupling introduces discrepancies between the demonstration and deployment settings. We address these discrepancies with deployment-aligned demonstration processing, which calibrates and temporally aligns platform motion with the deployed base response and masks and inpaints demonstrator hands in recorded observations. Compared with Mobile ALOHA and VR teleoperation, HALO enables up to 2.9x faster demonstration collection while maintaining comparable or higher success rates. Policies trained with HALO perform precise manipulation tasks, including zip-tie fastening, and achieve a 73.3% success rate on a six-stage long-horizon mobile manipulation task involving repeated coordinated base and arm motion. Project page: https://robbyant-research.github.io/Lingbot-ALOHA/
Shuai Yang, Fangyi Xu, Yishu Shen +2
The Hong Kong University of Science and Technology · Robbyant
While learned robotic policies hold promise for advancing generalizable manipulation, their practical deployment is often hindered by suboptimal execution speeds. Imitation learning policies are inherently limited by hardware constraints and the speed of the operator during data collection. In addition, there are no established methods for accelerating policies learned via imitation, and the empirical relationship between execution speed and task success remains underexplored. To address these issues, we introduce SpeedTuning, a reinforcement learning framework specifically designed to enhance the speed of manipulation policies. SpeedTuning learns to predict the optimal execution speed for actions, thereby complementing a base policy without necessitating additional data collection. We provide empirical evidence that SpeedTuning achieves substantial improvements in execution speed, exceeding 2.4x speed-up, while preserving an adequate success rate compared to both the original task policy and straightforward speed-up methods such as linear interpolation at a fixed speed. We evaluate our approach across a diverse set of dynamic and precise tasks, including pouring, throwing, and picking, demonstrating its effectiveness and robustness in enhancing real-world robotic manipulation. Videos and code are available at https://daivdyuan.github.io/speed-tuning/