LARR: A Leg-Arm Reuse Robot for cost-effective mobility and manipulation

Humanoid robots advance embodied intelligence but require over 20 degrees of freedom (DoFs), leading to high complexity and cost. This paper presents the Leg-Arm Reuse Robot (LARR), which adopts a novel design philosophy of reusing a single pair of 6-DoF commercial arms in a time-sharing manner for both mobility and manipulation, reducing active DoFs to only 13. To realize this concept, we first conducted human motion capture experiments to extract bio-inspired locomotion modes and quantify workspace requirements. The core mechanical challenge lies in designing a torso structure that simultaneously meets the workspace, strength, and lightweight demands of both locomotion and manipulation. We address this through a coherent methodology. The torso geometry is co-optimized using a QP-based reachability metric to maximize task coverage. Subsequently, an Experience-Based Structural Optimization method fed by dynamic simulation loads is applied, achieving over 33% mass reduction in key components while ensuring structural strength. Experiments show the LARR prototype traverses a 12 cm gap, an 8 cm step, and a 10 degree slope, and performs coordinated bimanual manipulation. The results confirm functional completeness and terrain adaptability comparable to traditional platforms, while substantially reducing joint-module hardware provision.