PaperlayerView exact revision on arXiv

Source-linked AI summary

Structure-Actuation Co-Design of All-Electric Heavy-Duty Robotic Manipulators

Mohammad Bahari, Amir Hossein Barjini, Jouni Mattila

arXiv:2610.36996v1cs.ROcs.SYeess.SY

TL;DR

The paper addresses the coupled design of heavy-duty manipulators, where actuator mass affects robot dynamics and structural loading. It combines mechanical optimization, discrete actuator searches, system-level package coordination, and mass feedback; the nominal study returns eight feasible designs and shows that one-pass sizing changes the system trade-off.

  • Problem

    Manipulator geometry and actuator packages jointly determine mechanical performance and actuator demand, while actuator package mass also changes robot dynamics and structural loading.

    Source: §I · p. 1

  • Method

    A mixed discrete–continuous bilevel framework optimizes mechanical geometry and topology, searches feasible discrete actuator packages, coordinates them by system objectives, and feeds their masses back into the robot model.

    Source: §I · p. 1

  • Results

    Eight feasible nondominated designs satisfy all 12 prescribed SE(3) poses, and one-pass sequential sizing changes hardware mass by up to 2.13%, co-design cost by 10.20%, and supply energy by 9.40% in magnitude.

    Source: §V · p. 6

  • Takeaways & Limitations

    Actuator-mass feedback can alter the selected discrete package and resulting system trade-off, while payload screening shows increasing mass–cost–energy demand and topology changes with carried load.

    Source: §V · p. 6

    The reported solutions are a fixed-budget nondominated approximation rather than a globally converged Pareto front, and the study uses engineering structural and motor-loss models, a fixed-design wrist, and omits wrist-actuator energy.

    Source: §V · p. 6

Abstract

from arXiv · show

This paper presents a mixed discrete-continuous multiobjective bilevel framework for the structure-actuation co-design of all-electric heavy-duty robotic manipulators. The upper level optimizes manipulator geometry and actuation topology, while joint-specific lower-level searches construct feasible archives of discrete motor, gearbox, and screw configurations. A system-level coordinator selects actuator packages according to hardware mass, co-design cost, and loss-aware supply energy, and the resulting actuator masses are fed back to the robot model until a mass-consistent fixed point is obtained. Evaluation on a full-pose SE(3) benchmark with a 100-kg tool payload yields eight feasible nondominated designs, all satisfying the 12 prescribed poses. Actuator-mass coupling converges in three iterations for every reported design. A fixed-design ablation shows that one-pass sequential sizing changes hardware mass by up to 2.13%, co-design cost by 10.20%, and supply energy by 9.40% in magnitude. Payload screening further reveals increasing mass-cost-energy demand and changes in competitive mechanism topology with carried load.

I. INTRODUCTION

All-electric heavy-duty manipulators require mechanical architecture and actuator packages to be designed together because actuator mass changes robot dynamics and structural loading. The paper proposes a bilevel co-design framework and evaluates it on a full-pose manipulator task.

I. INTRODUCTION

  • Actuator package mass modifies robot dynamics and structural loading, making geometry and actuation a coupled design problem rather than sequential sizing stages.Heavy-duty manipulators also require high payload capacity, extended reach, and reliable operation under demanding loads.Source: §I · p. 1
  • The framework jointly optimizes continuous mechanical geometry and mechanism topology with discrete actuator packages, coordinating them by hardware mass, co-design cost, and loss-aware supply energy.Actuator masses feed back into the robot models until a mass-consistent fixed point is reached.Source: §I · p. 1

A. Architecture and Task

  • The mechanical design uses 11 variables for link dimensions, mechanism choices, and linkage geometry; the benchmark prescribes 12 TCP positions and orientations.The closed dynamic-sizing trajectory tracks the TCP positions, while the full arm–wrist chain is checked for kinematic feasibility.Source: §I · p. 1, A. Architecture and Task · pp. 1–2, Figure 2 · p. 2
  • The main arm has three degrees of freedom, while a fixed three-degree-of-freedom X–Y–X spherical wrist supplies end-effector orientation and is excluded from co-design.Joint 1 is rotary; joints 2 and 3 may use rotary actuation or three-bar or four-bar mechanisms.Source: A. Architecture and Task · p. 1

B. Dynamics, Linkages, and Structural Screen

The dynamics and structural screen evaluate designs under actuator-dependent mass distributions and carried loads. Structural feasibility is assessed with a fixed-section axial-plus-bending engineering model, not a certification model.

B. Dynamics, Linkages, and Structural Screen

  • The main-arm dynamics include downstream link, actuator, wrist, and tool masses, with distal carried mass defined as wrist plus tool mass.The wrist contributes to hardware mass, while both wrist and tool contribute to dynamic and structural loading.Source: B. Dynamics, Linkages, and Structural… · p. 2
  • For EMLA-driven joints, effective linkage angle depends on joint type and linkage offset, and force estimation uses numerical regularization.Invalid installation geometries or insufficient actuator-length margins are rejected before dynamic evaluation.Source: B. Dynamics, Linkages, and Structural… · p. 2
  • All links use a fixed hollow-square steel section, so cross-section dimensions are not part of the optimized design.The section has outer width 0.150 m, wall thickness 0.005 m, and steel density 7850 kg/m3.Source: B. Dynamics, Linkages, and Structural… · p. 2
  • The axial-plus-bending structural check is an engineering feasibility screen, not a local-stress, fatigue, or buckling certification model.The screen uses downstream link, actuator, wrist, and tool masses in the structural loading.Source: B. Dynamics, Linkages, and Structural… · p. 2

C. Discrete Actuator Packages and Loss-Aware Energy

Joint-level actuator selection searches discrete catalogue packages subject to actuator and topology constraints. A loss-aware surrogate estimates supply energy, with explicit exclusions that limit its scope.

C. Discrete Actuator Packages and Loss-Aware Energy

  • Each joint’s discrete package selects a catalogue motor and gearbox ratio, plus a screw lead for an EMLA.Motor torque, speed, power, and current limits are enforced; EMLA feasibility also includes screw strength, buckling, and critical speed.Source: C. Discrete Actuator Packages and… · p. 2
  • Supply energy is estimated with a catalogue-calibrated, load-dependent loss surrogate using positive net motor-plus-loss power.Negative net power is not credited as regeneration.Source: C. Discrete Actuator Packages and… · pp. 2–3
  • The energy model supports comparative actuator selection but does not explicitly resolve speed-dependent iron, windage, inverter, or thermal losses.Wrist-actuator energy is also excluded because the timed dynamic model covers only the co-designed main arm.Source: C. Discrete Actuator Packages and… · p. 3

III. MASS-CONSISTENT BILEVEL CO-DESIGN

Mechanical design is the leader decision, and actuator selection is its mass-consistent follower response. Joint-level searches retain feasible alternatives for system-level coordination rather than choosing each actuator independently.

III. MASS-CONSISTENT BILEVEL CO-DESIGN

  • Each mechanical design is evaluated under the mass distribution induced by its converged actuator packages, making actuator selection the follower response to the leader design.The follower response is obtained through joint-level search, system-level coordination, and actuator-mass feedback.Source: §III · p. 3

A. Joint-Level Actuator Search

  • For a given mechanical design and actuator-mass vector, robot simulation supplies the joint histories used for actuator selection.The lower-level objectives are family-dependent.Source: A. Joint-Level Actuator Search · p. 3
  • The lower-level feasible follower set is subject to actuator and topology-dependent constraints, and each joint search retains a bounded archive rather than one solution.Nondominated solutions are prioritized while additional feasible alternatives remain available for system-level coordination.Source: A. Joint-Level Actuator Search · p. 3

B. System-Level Package Coordination and Mass Feedback

Joint actuator archives are coordinated as realizable three-joint packages, and the selected actuator masses are fed back into the robot model. This makes upper-level objectives depend on mass-consistent installed actuators.

B. System-Level Package Coordination and Mass Feedback

C. Upper-Level Objectives and Feasibility

The upper-level objectives account for link, actuator, and wrist hardware mass, while feasibility requires a mass-consistent design that satisfies the prescribed task and physical constraints.

C. Upper-Level Objectives and Feasibility

IV. RESULTS AND DISCUSSION

The numerical setup evaluates a 100 kg tool payload with full-pose feasibility and dynamic and structural limits. It also specifies a common cost reference and a reproducible, fixed-resolution nominal campaign.

A. Numerical Setup

  • The nominal study combines a 100 kg tool payload with a fixed 50 kg spherical wrist, yielding 150 kg of distal carried mass.Source: A. Numerical Setup · p. 3
  • Full-pose feasibility requires all 12 prescribed SE(3) targets, at least 5° joint-limit and wrist-singularity margins, and a main-arm condition number no greater than 100.Source: A. Numerical Setup · pp. 3–4
  • For each mechanical design, the workflow simulates joint demands, constructs topology-dependent joint archives, selects a package, updates masses, and performs a final mass-consistent evaluation.Source: A. Numerical Setup · p. 4
  • Dynamic tracking error is limited to 20 mm and structural stress to 250 MPa.Source: A. Numerical Setup · p. 4
  • The cost index uses the lowest co-design cost across the nominal and payload-screening datasets as a common reference; the nominal campaign uses full simulation resolution and fixed random seeds.Source: A. Numerical Setup · p. 4

B. Nominal 100-kg Tool-Payload Co-Design

The nominal search finds feasible nondominated structure–actuation designs with trade-offs across hardware mass, co-design cost, and supply energy. Payload screening shows increasing minimum observed demand and load-dependent competitive topologies.

B. Nominal 100-kg Tool-Payload Co-Design

  • 320 upper-level designs yielded eight feasible nondominated solutions satisfying all kinematic, dynamic, structural, actuator, and mass-consistency requirements.Source: B. Nominal 100-kg Tool-Payload Co-Design · p. 4
  • The eight-design set spans 387.62–419.74 kg hardware mass, 1.187–1.310 normalized co-design cost, and 30.53–36.90 kJ supply energy, with improvement in one objective generally requiring sacrifice in another.Source: B. Nominal 100-kg Tool-Payload Co-Design · p. 4
  • DM and DE share 4-bar/3-bar topology and principal dimensions but use different J1/J2 motor packages, while DC uses 4-bar/4-bar and 2.2-kW motors at both linearly actuated joints.Source: B. Nominal 100-kg Tool-Payload Co-Design · p. 4

C. Payload Sensitivity

  • From 20 to 140 kg tool payload, minimum observed hardware mass rises from 348.87 to 442.16 kg, cost index from 1.000 to 1.394, and supply energy from 22.86 to 40.33 kJ.Source: C. Payload Sensitivity · p. 4
  • Competitive topologies shift with load: 20 kg has 3-bar/4-bar and 4-bar/4-bar; 60 and 100 kg add 4-bar/3-bar; 140 kg retains 3-bar/4-bar and 4-bar/3-bar.Source: C. Payload Sensitivity · p. 4

D. Effect of Actuator-Mass Feedback

The fixed-design ablation compares one-pass actuator sizing with mass-coupled reselection. Reselection changes objective values and can change the selected discrete packages.

D. Effect of Actuator-Mass Feedback

  • The one-pass baseline sizes actuators from the common initial estimate, freezes the selected package, and reevaluates it using installed masses without reselection.The ablation compares this procedure with the fully coupled solution.Source: D. Effect of Actuator-Mass Feedback · pp. 4–5
  • Hardware mass changes by 0.43–2.13%, co-design cost by −3.81 to +10.20%, and supply energy by −8.80 to +9.40% after actuator reselection.These are the one-pass baseline differences from the fully coupled solution.Source: Table III · p. 5, D. Effect of Actuator-Mass Feedback · p. 5
  • DM changes its J2 screw lead from 10 to 7 mm/rev after coupling, while DE changes both its J1 package and J2 screw selection.The discrete follower response changes along with the realized trade-off.Source: D. Effect of Actuator-Mass Feedback · p. 5
  • DM’s mass iteration starts at T kg, updates to [77.36, 43.72, 21.49]T kg, then to [77.86, 41.40, 21.49]T kg; the latter is reproduced on the next iteration.The remaining reported nominal designs meet the mass-convergence criterion in three iterations.Source: D. Effect of Actuator-Mass Feedback · pp. 5–6

E. Constraint and Physical Validation

All eight nominal designs satisfy the prescribed physical constraints and 12 full-pose targets, and actuator-mass coupling converges in three iterations. The reported set remains a fixed-budget approximation, with additional model and wrist scope limitations.

E. Constraint and Physical Validation

  • All eight nominal designs satisfy the physical constraints and 12 full-pose targets; worst-case tracking error is 9.73 mm and structural stress is 128.35 MPa, below limits of 20 mm and 250 MPa.Mass coupling converges in three iterations with zero final residual and no detected cycle.Source: E. Constraint and Physical Validation · p. 6, §V · p. 6
  • The finite-budget results are a nondominated approximation, not a globally converged Pareto front; limitations also include engineering models, a fixed-design wrist, and omitted wrist-actuator energy.Future work includes higher-fidelity component and thermal models, robust co-design under uncertainty, and experimental validation.Source: E. Constraint and Physical Validation · p. 6, §V · p. 6

V. CONCLUSION

  • The framework jointly optimizes mixed mechanism topology and continuous geometry with discrete actuator packages, using system-level coordination and actuator-mass feedback.Source: §V · p. 6
  • One-pass sequential sizing changes hardware mass by up to 2.13%, co-design cost by 10.20%, and supply energy by 9.40% in magnitude.The ablation also shows that feedback can alter the selected discrete package and resulting system trade-off.Source: §V · p. 6
  • Payload screening shows increasing mass–cost–energy demand and changes in competitive mechanism topology with carried load.Source: §V · p. 6
Loading 2610.36996v1…