With the rapid advancement of global aerospace technology, human space exploration is progressively extending from near-Earth orbit to the lunar sphere. Large-scale deep-space exploration activities, such as the construction of lunar research stations and lunar bases, are gradually becoming a reality. Major spacefaring entities, including China, the United States, Europe, and Japan, have successively proposed their own lunar exploration and long-term residency programs. Current lunar exploration missions now encompass complex scenarios such as scientific investigation, in-situ resource utilization, lunar construction, and operational maintenance, which impose entirely new demands on lunar robotic technologies. However, existing lunar and Mars exploration missions still predominantly rely on pre-programmed commands and ground-based teleoperation, exhibiting limited intelligence levels. They face significant challenges in addressing five core issues: multi-task high-precision manipulation, autonomous navigation and obstacle avoidance, self-learning interactive collaboration, adaptation to extreme environments, and high-reliability long-duration operation. Consequently, the systematic integration of emerging technologies—including artificial intelligence, deep learning, large-scale models, and embodied intelligence—into lunar robotic design, so as to achieve the transition from programmed robots to intelligent robots, has become a critical technological bottleneck for supporting future large-scale lunar exploration and development.
In a recent study published in Space: Science & Technology, the research team led by Wang Xiaowei from the China Academy of Launch Vehicle Technology systematically elucidated the developmental requirements and research progress of lunar intelligent robotics technology. Based on the phased milestones of human lunar exploration, the study first proposes a developmental roadmap for lunar robots, categorizing them into two major stages: the programmed era and the intelligent era, with the latter further subdivided into three hierarchical levels—weak intelligence, intelligence, and general intelligence. To address the primary technical challenges posed by complex lunar surface operational scenarios, the study introduces a design philosophy centered on "intelligence + new energy" empowerment and an architecture comprising a "brain + cerebellum + body." A comprehensive technological framework is established, encompassing key technologies such as environmental perception, autonomous decision-making, large-model reasoning, intelligent navigation, autonomous obstacle avoidance, and precision manipulation. The study systematically elaborates on the key technologies in three major directions: the "brain" (perception, decision-making, reasoning), the "cerebellum" (navigation, obstacle avoidance, motion control), and the "body" (actuation, manipulation, power, environmental adaptation). For representative key technologies—including autonomous navigation and path planning, multi-arm collaborative compliant control, and high-mobility multi-modal locomotion mechanisms—prototype verification has been conducted. Preliminary experiments demonstrate that the visual perception algorithm can achieve high-precision mapping and localization under complex illumination conditions, and that the wheel–leg hybrid locomotion mechanism can effectively climb slopes exceeding 20° and overcome obstacles higher than the wheel diameter. This research provides a systematic reference framework for the planning, technological development, and engineering application of intelligent robotic technologies in the construction of lunar research stations and lunar bases, thereby laying an important foundation for establishing an intelligent robotic technology system oriented toward future large-scale lunar exploration.
First, this paper systematically analyzes the evolutionary trends of lunar exploration missions and their associated requirements for intelligent robotic technologies. As human space exploration shifts from near-Earth space to the lunar domain, large-scale deep-space activities such as the construction of lunar research stations and lunar bases are progressively materializing. China and Russia are jointly advancing the International Lunar Research Station (ILRS) program, the United States is implementing the Artemis program, and Europe and Japan have respectively put forward visions for a Moon Village and a lunar industry. Lunar exploration missions have evolved from simple unmanned scientific surveys to encompass diverse complex scenarios, including *in-situ* resource utilization, lunar construction, and operational maintenance. The capability demands on robotics exhibit a progressive pattern across different mission phases: from basic mobility and single-arm manipulation during the unmanned scientific exploration phase, to multi-task operations and autonomous manipulation during the unmanned lunar research station phase, then to complex assembly, construction, and human–robot collaboration during the lunar base phase, and ultimately to fully intelligent autonomous operations during the lunar community phase. Based on this evolutionary trajectory of missions, the paper proposes a two-stage roadmap for lunar robot development: the programmed robot era, characterized by pre-programmed commands and teleoperation, corresponding to the unmanned scientific exploration phase; and the intelligent robot era, which is further subdivided into three hierarchical levels—weak intelligence, intelligence, and general intelligence—corresponding respectively to the unmanned research station, lunar base, and lunar community phases. Within the framework of the "brain" architecture, the study addresses core technological issues including environmental perception, mission planning and decision-making, autonomous obstacle avoidance, health monitoring, and human–robot collaboration.
Second, this paper systematically constructs a technological framework for lunar intelligent robotics from three dimensions, namely, the "brain," the "cerebellum," and the "body." As illustrated in Fig. 1, this framework defines the "brain" as the core command center of the robot, responsible for receiving external information, planning actions, and generating specific instructions. It encompasses key technologies such as multimodal perception and information fusion, autonomous mission planning and decision-making, large-model reasoning, path planning and obstacle avoidance, health monitoring, cloud computing and intelligent chips, as well as human–robot and swarm collaboration. The "cerebellum" is responsible for fine-grained regulation of motion control, receiving commands from the "brain" and adjusting actions in real time based on the current motion state; it includes technologies for autonomous navigation and localization in complex terrains, multi-arm collaborative compliant control, and reinforcement learning–based motion control. The "body" serves as the executor of commands, possessing the strength and flexibility required to accomplish various complex maneuvers; it incorporates technologies such as high-torque long-life modular joints, high-mobility multi-modal locomotion mechanisms, end effectors, high-specific-energy distributed power systems, wireless power transmission, and environment-adaptive design. To address the challenges of high-precision mapping and localization in the lunar south pole region, characterized by complex illumination and low-texture environments, the research team developed a visual perception algorithm based on a lightweight deep network. As shown in Fig. 2, this algorithm achieves stable feature extraction and matching under varying illumination conditions and accurately recovers depth information in scenes with repetitive textures, thereby laying a foundation for autonomous navigation on the lunar surface.
Finally, this paper conducts ground-based prototype verification for representative key technologies. In terms of multi-arm collaborative compliant control, as shown in Fig. 3, the research team established a ground verification platform comprising an equivalent manipulator, an end quick-change mechanism, a controller, and a multifunctional tool kit. A nonlinear compliant control method was employed to achieve high-precision position tracking and contact force buffering for the end effector. Through 50 repeated measurements of position accuracy and 30 repeated measurements of orientation accuracy, the locking precision of the end quick-change device in all three directions was verified to meet the requirements for fine manipulation. In terms of the high-mobility multi-modal locomotion mechanism, as illustrated in Fig. 4, the wheel–leg hybrid mobility subsystem adopts four independent deployable mechanisms, each chain possessing three degrees of freedom. In the vehicle body lifting experiment, the robot stably completed the chassis elevation from a squatting posture through coordinated wheel–leg motion within approximately 10 seconds; the motor current and torque feedback remained within safe operating ranges, with no jamming at the joint pivots. The mechanism effectively climbed a 20.2° slope and traversed a 28.3 mm step (exceeding twice the wheel diameter), demonstrating favorable adaptability to unstructured obstacles. Based on the above analyses, the paper proposes three development recommendations: first, to establish a technical framework and a consensus on technology classification for lunar intelligent robotics, clearly defining the functional scope and technical metrics for each level; second, to pursue a phased advancement strategy, with near-term breakthroughs in weak-intelligence technologies to build engineering application capabilities, and medium-to-long-term development of intelligence and general-intelligence technologies; and third, to formulate international industry standards encompassing overall system design, interface specifications, and environmental adaptability, thereby facilitating interoperability and collaborative development among robots from different nations. This study provides a systematic reference framework for the planning, technological development, and engineering application of intelligent robotic technologies in the construction of lunar research stations and lunar bases.
