In this paper, we propose ThinkLocoManip, an agentic harness for legged loco-manipulation designed to tackle complex, long-horizon tasks in large-scale environments. Existing approaches typically follow either direct action generation or a plan-then-act paradigm, both of which can struggle with the generalization and error-recovery capabilities required for extended task execution. In contrast, ThinkLocoManip adopts an interleaved reasoning-and-acting paradigm, allowing the agent to continuously reason about its current state, execute actions, observe their outcomes, and adapt its strategy accordingly. Building on this paradigm, our agentic harness orchestrates a suite of specialized tools for perception, navigation, manipulation, and locomotion, enabling flexible composition of heterogeneous capabilities. Unlike existing robotic agentic harnesses, ThinkLocoManip explicitly incorporates locomotion as a tool domain for agent reasoning and execution. We further design a set of skills that explicitly account for both tool usage within each functional domain and coordination across domains, supporting effective execution of complex behaviors over long horizons. To tackle task failure during long-horizon interactions, we develop keyframe thinking mechanism to preserve critical execution context and support efficient error recovery. Extensive real-world experiments demonstrate that ThinkLocoManip substantially outperforms representative methods based on both direct action generation and plan-then-act paradigms across challenging long-horizon legged loco-manipulation tasks. Notably, ThinkLocoManip achieves a success rate of 77%, more than twice that of the strongest plan-then-act baseline (37.1%). Our experiments further validate the effectiveness of interleaved reasoning and acting, tool orchestration, and coordinated skill execution.
Figures & tables
Fig. 1: (a) An overview of our method for indoor and in-the-wild loco-manipulation tasks. (b) Performance comparison of long-horizon real-world tasks.
Fig. 2: The Comparison of different frameworks for legged loco-manipulation . (a) direct action generation method uses an end-to-end policy to map observations directly to robot actions. (b) plan-then-act method constructs a tool invoking plans and then execute each tool. (c) Our approach design an agentic harness, enabling interleaved reasoning-and-acting.
Fig. 3: The overall framework of our method . (a) The agent performs task planning to understand task instruction and decompose it into several sub-tasks. (b) ThinkLocoManip orchestrates a suite of Primitive Tools for perception, navigation, manipulation, and locomotion. We also design ThinkLocoManip-Skill that explicitly accounts for both tool usage within each functional domain and coordination across domains. (c) Keyframe Thinking mechanism enables error recovery from task failure. (d) The whole execution workflow of our method.
Primitive Tool
Function
Perception
Capture Image
Take RGB/Depth Image using camera.
Prompt Detection
Detect the object matching the prompt.
Navigation
Map-based Nav
Navigate to waypoint based on 3D Map.
Target-based Nav
Navigate to (x,y,z) in the robot’s base frame.
TABLE I: Library of Primitive Tools.
Fig. 4: The system setup and execution process of five real-world indoor tasks . (a) We demonstrate the robot and objects used in the experiment. (b)-(f) We conduct experiments on five indoor tasks, illustrating the complete execution process from navigation to task completion and demonstrating the effectiveness of our agentic harness framework.
Method
Table Grasping
Ground Grasping
Button Pressing
Chest Manip.
Cabinet Manip.
Avg.
Succ
TaskC (%)
Succ
TaskC (%)
Succ
TaskC (%)
Succ
TaskC (%)
Succ
TaskC (%)
SuccR (%)
TaskC (%)
ACT
4/20
40.0
5/20
43.8
0/10
33.3
3/10
44.0
1/10
28.0
18.6
37.8
π0.5
6/20
47.5
6/20
47.5
0/10
33.3
3/10
47.5
2/10
40.0
24.3
43.2
plan-then-act
10/20
62.5
7/20
51.3
4/10
60.0
2/10
48.0
3/10
56.0
37.1
55.6
Ours
17/20
88.8
18/20
92.5
8/10
86.7
6/10
72.0
5/10
66.0
77.0
81.2
TABLE II: Performance comparison on real-world long-horizon loco-manipulation tasks. We color the best and second best results. The improvement in average success rate over the second-best method is shown in parentheses.
Long-horizon humanoid loco-manipulation requires composing versatile whole-body skills and reliable high-level decision making. Existing methods often coordinate pretrained skills with scripted planners, finite-state machines or task-specific model-free policies, restricting their ability to handle complex task sequences. To address this limitation, we propose LUCID, a hierarchical model-based reinforcement learning framework that plans over reusable skills through imagined rollouts of a learned dynamics model. LUCID first trains a structured latent-conditioned low-level policy via adversarial imitation and then freezes it while jointly learning a high-level policy and macro-dynamics world model. The world model predicts the temporally extended state transitions induced by latent decisions, enabling high-level policy optimization through imagined rollouts. We evaluate our framework across various simulated multi-object rearrangement scenarios. Experimental results show that LUCID improves the full-task success and partial-completion rates compared to prior baseline methods, demonstrating its effectiveness in complex sequential loco-manipulation tasks.
Cheng Guo, Mingzhe Ni, Angelo Cangelosi +1
Department of Computer Science, University of Manchester, Manchester, UK · Human-Robot Interfaces and Interaction Laboratory, Italian Institute of Technology, Genoa, Italy
Long-horizon mobile manipulation presents significant challenges due to compounding execution errors and capacity interference between locomotion and arm control. While recent Vision-Language-Action models excel at short-horizon tasks, they lack the hierarchical reasoning required for multi-stage objectives. Furthermore, existing hierarchical agents suffer from rigid sub-task mapping, inflexible replanning, and a lack of continuous learning. To address these limitations, we introduce MobiAgent, a dual-loop agentic framework that bridges robust deployment execution and recursive policy self-improvement. During deployment, the Inner Loop decouples high-level reasoning from low-level control through highly composable atomic skills. It employs Vision-Language models for receding-horizon planning and visual reflection, dynamically composing skills to ensure robust error recovery. These skills are executed by specialized flow-matching experts that share a unified VLM backbone, maximizing reusability while mitigating capacity interference. Concurrently, the Outer Loop drives automated lifelong learning by autonomously segmenting and verifying deployment rollouts, clustering them to discover atomic skills, and continuously fine-tuning the skill library without human annotations. Evaluations on RoboCasa, BEHAVIOR-1K, and real-world tasks demonstrate the effectiveness of MobiAgent. It outperforms π0.5-TA by 22.5 percentage points on BEHAVIOR-1K and enables robust recovery from execution failures. Through autonomous data recycling, success improves from 7.50% to 27.50% on RoboCasa and from 32.5% to 57.5% on Astribot S1.
Chenzhi Liu, Yue Zhang, Jiehong Lin +4
The University of Hong Kong · Astribot · Southern University of Science and Technology
Recent vision-language-action (VLA) systems have demonstrated strong capabilities in embodied manipulation. However, most existing VLA policies rely on limited observation windows and end-to-end action prediction, which makes them brittle in long-horizon, memory-dependent tasks with partial observability, occlusions, and multi-stage dependencies. Such tasks require not only precise visuomotor control, but also persistent memory, adaptive task decomposition, and explicit recovery from execution failures. To address these limitations, we propose a dual-system framework for long-horizon embodied manipulation. Our framework explicitly separates high-level semantic reasoning from low-level motor execution. A high-level planner, implemented as a VLM-based agentic module, maintains structured task memory and performs goal decomposition, outcome verification, and error-driven correction. A low-level executor, instantiated as a VLA-based visuomotor controller, carries out each sub-task through diffusion-based action generation conditioned on geometry-preserving filtered observations. Together, the two systems form a closed loop between planning and execution, enabling memory-aware reasoning, adaptive replanning, and robust online recovery. Experiments on representative RMBench tasks show that the proposed framework substantially outperforms representative baselines, achieving a 32.4% average success rate compared with 9.8% for the strongest baseline. Ablation studies further confirm the importance of structured memory and closed-loop recovery for long-horizon manipulation.
Zhen Liu, Xinyu Ning, Zhe Hu +8
Beijing University of Posts and Telecommunications · InspireOmni AI · Tsinghua University