PaperlayerView exact revision on arXiv

Source-linked AI summary

OmniQuad: a Wheeled-Legged Hybrid Robot with Omnidirectional Wheels

Francesco Iotti, Alok Ranjan, Franco Angelini, Manolo Garabini

arXiv:2610.38502v1cs.RO

TL;DR

Wheeled and legged robots have complementary strengths, but wheels face obstacle limits and legs may be slower or less efficient. OmniQuad integrates mecanum wheeled-feet with legs and evaluates the robot in indoor and outdoor tasks. It reaches 2 m/s forward and 1.5 m/s laterally without slippage, while lateral motion has greater friction and vibration.

  • Problem

    Wheels can be efficient but struggle with obstacles, while legged robots can traverse rugged terrain but may be slower and less efficient.

    Source: §1 · p. 2

  • Method

    OmniQuad combines mecanum wheels attached to its feet with legs and a control system grounded in both locomotion modes.

    Source: §6 · p. 24

  • Results

    2 m/s forward and 1.5 m/s lateral speeds were reached without slippage, though lateral motion showed higher friction and vibration.

    Source: §5.2.3 · p. 14

  • Takeaways & Limitations

    The integrated robot was demonstrated navigating static obstacles in structured environments using encoder-based odometry and orientation data.

    Source: §6 · p. 24

    Current autonomy is limited to fundamental navigation in structured environments with static obstacles, using encoder-based odometry and orientation data.

    Source: §6 · p. 24

Abstract

from arXiv · show

Combining wheeled and legged locomotion within a single robotic platform offers the potential to harness the advantages of both systems. The development of such hybrid systems remains an active area of research. While legged robots exhibit remarkable adaptability for traversing uneven and rugged terrains, their relatively low speed limits their practicality in indoor environments. Conversely, wheeled robots demonstrate higher efficiency on flat surfaces but often encounter difficulties when navigating obstacles such as steps. This paper introduces OmniQuad, a novel hybrid robot that integrates mecanum wheels with legged locomotion to exploit the benefits of both modalities. Mecanum wheels provide omnidirectional mobility, enhancing the robot's manoeuvrability by enabling agile movement in confined spaces and the ability to maintain a consistent orientation during task execution. In this work, we first present the design of a custom wheeled-foot featuring mecanum wheels. Subsequently, we describe the complete robot assembly, comprising these wheeled feet and EM-Act actuators. Finally, we detail the control scheme developed to govern the OmniQuad and evaluate its performance through experimental trials conducted in both indoor and outdoor environments.

1. Introduction

Wheeled and legged robots each have useful but distinct locomotion strengths, motivating a hybrid that combines omnidirectional maneuverability with obstacle traversal. OmniQuad integrates mecanum wheels and legs and is evaluated in indoor and outdoor tasks.

1. Introduction

  • Mecanum wheels support maneuverability and high load capacity, but wheel friction, vibration, and slippage can reduce efficiency, especially beyond flat terrain.The paper also notes efforts to mitigate these effects with springs.Source: §1 · pp. 1–2
  • Wheeled robots offer speed and efficiency but struggle with obstacles, while quadrupeds traverse rugged terrain but may be slower and less efficient.The paper frames hybrid locomotion as a way to combine the strengths of both modes.Source: §1 · p. 2
  • The paper presents OmniQuad’s mecanum wheeled-foot, integrated robot and control system, then evaluates obstacle navigation, fixed-orientation maneuvers, and an outdoor grid-pattern survey.The tasks include passing beneath or scaling over obstacles.Source: §1 · pp. 2–3

2. Literature Review

Prior wheeled-legged designs address different locomotion needs but often trade off obstacle traversal, locomotion efficiency, or mechanical simplicity. OmniQuad combines mecanum wheels and legs for omnidirectional mobility and uneven-terrain navigation.

2. Literature Review

  • Transformable-wheel robots can climb obstacles, but reported designs face alignment, obstacle-size, or adaptability constraints.The cited passive design can tackle stairs, but may be less adaptable on highly uneven or complex terrain.Source: §2 · p. 3
  • Several wheeled-legged designs do not combine wheeled and legged motion simultaneously, while some one-joint-per-leg systems cannot lift their legs to clear obstacles.Other designs trade locomotion efficiency or mechanical simplicity for their capabilities.Source: §2 · p. 3
  • Other systems include tracked-leg designs, two-leg/two-wheel robots, and specialized wheel platforms, each with stated limits in indoor use, obstacle traversal, steering, or intended role.For example, Swiss-mile uses regular wheels that can make steering challenging.Source: §2 · p. 4
  • OmniQuad integrates mecanum wheels and legs to maintain body orientation without relying exclusively on leg rotation, and uses fewer wheel degrees of freedom than Centauro and Momaro.Its legged mode supports uneven-terrain travel, obstacle climbing, and height adjustment beneath obstacles.Source: §2 · p. 4

3. OmniQuad Design and Electronics

OmniQuad’s design combines mecanum wheeled-feet, a four-legged chassis, and onboard electronics. The design choices and assembly support wheel-based omnidirectional movement alongside legged locomotion.

3.1. Actuation System

  • Mecanum wheels were selected for omnidirectional movement based on implementation, mounting, and actuation considerations, with mostly plastic wheels chosen to limit weight.Wheel diameters were analyzed for stability and total torque.Source: §3.1 · p. 4
  • The actuation design uses an Antigravity MN4004 brushless motor and an 80 mm mecanum wheel, with a two-stage AASUN AM-28P planetary gearbox at a 15:1 ratio.The selected wheel yields a stated total torque of 10.5 Nm; the gearbox is chosen for compactness and high gear ratio.Source: §3.1 · p. 5

3.2. Design of the Wheeled-Foot

  • The wheeled-foot aligns its motor, gearbox, and mecanum wheel on one axis; its link encloses the gearbox and the assembly weighs 0.42 kg.The assembly measures 230 mm long and 150 mm wide, and includes a bearing and cable-protecting casing.Source: §3.2 · p. 5

3.3. OmniQuad

  • Each of OmniQuad’s four legs has two EM-Act actuators and a wheeled-foot with a third actuator, while mecanum wheels on 2-DOF legs enable lateral movement.The legs mount to a cross-linked aluminium chassis.Source: §3.3 · pp. 5–6

3.4. Electronics

  • OmniQuad uses twelve motors, with a Raspberry Pi 4 and Pi3hat providing central control, low-level joint control, and sensor-data acquisition.The leg motors connect through per-leg daisy chains; each leg links to a Pi3hat channel.Source: §3.4 · pp. 7–8

4. Modelling and Control

The section develops OmniQuad’s wheel and leg kinematics, hybrid locomotion and dynamic model, then describes the control system and its obstacle-handling modules.

4.1. Omnidirectional Vehicle

  • Four independently actuated Mecanum wheels, arranged with matching roller axes along the diagonals, provide forward, lateral, and rotational motion through their combined force vectors.The wheel velocities govern the platform’s linear and angular velocity.Source: §4.1 · pp. 8–9

4.2. Legged System

  • Two revolute joints let each leg move the base in three dimensions, raise a foot independently for obstacle crossing, and adjust base height when all legs move together.Coordinated front and back leg motion also supports pitch adjustment when only three legs contact the ground.Source: §4.2 · pp. 9–10

4.3. Hybrid System

  • Wheel-to-leg transitions depend on keeping all four Mecanum wheels in ground contact, while the legs position the wheels and absorb impacts as suspension.When approaching an obstacle edge, the controller moves the wheel forward beyond its radius so contact aligns with the wheel center and gravity direction.Source: §4.3 · p. 10

4.4. Dynamic Model

  • The total dynamic model integrates quadruped and omnidirectional-vehicle dynamics, adding wheel angular positions to the robot’s combined state-space model.The quadruped is modeled as a floating-base system; the vehicle model treats it as a planar rigid body.Source: §4.4 · pp. 10–11

4.5. Control System

  • The control system derives inverse kinematics, detects and handles obstacles, and coordinates robot motion through a full control diagram.The overall motion objective is to keep the centre of mass within the support polygon while following a predetermined trajectory.Source: §4.5 · p. 11, §4.5.3 · p. 12

4.5.1. Inverse Kinematics module

  • Leg inverse kinematics maps a desired foot position to joint angles, with the sign choice for q2 selecting a C-shaped or X-shaped gait.The wheel inverse kinematics takes the base’s linear and angular velocity as input.Source: §4.5.1 · p. 11

4.5.2. Obstacle Detection and Obstacle Climbing modules

  • Wheel-current torque measurements detect obstacles and estimate their height, while the climbing module uses support-polygon planning and IMU readings to climb or avoid them.For a stuck wheel, lifting continues while the wheel rotates until torque drops near zero; when two wheels are stuck, the robot switches to legged climbing.Source: §4.5.2 · p. 12

4.5.3. Control Diagram

  • The control diagram combines trajectory planning, inverse kinematics, obstacle-module updates, and motion tracking; a ROS2 hardware interface actuates joints through FD-CAN.A gamepad can set desired linear and angular velocities and the floating-base height adjustment rate.Source: §4.5.3 · pp. 12–13, Figure 8 · p. 12

5. Experimental Validation

The experiments evaluated OmniQuad’s wheel-foot actuation, locomotion across indoor and outdoor terrain, obstacle handling, and hardware robustness. Wheel motion was faster forward than laterally, with lateral friction and vibration limiting performance; the robot also climbed a 20° slope at 0.1 m/s.

5. Experimental Validation

  • The experiments tuned joint PD controllers, tested wheel-foot performance, examined obstacle traversal and outdoor joystick control, and included obstacle-detection, durability, and leg-payload stress tests.Leg-joint gains were 150 Nm/rad proportional and 0.5 Nms/rad derivative; wheel joints used a 0.95 Nms/rad derivative gain.Source: §5 · p. 13, §5.1 · p. 13

5.2.1. Experimental setup

  • Outdoor wheel-foot tests covered asphalt, grass, soft polyurethane foam, and a sloped surface, alongside indoor forward and lateral movement tests.The test setup assessed slippage, friction, and actuation-system behavior across terrains.Source: §5.2.1 · pp. 13–14, Figure 10 · p. 14, Figure 12 · p. 14

5.2.2. Test description

  • Wheel tests used an acceleration, constant-velocity, deceleration, and reverse trajectory in both forward and lateral directions; outdoor tests reached 1 m/s forward and -1 m/s in reverse.The slope test used a 0.5 m/s forward reference and the same reverse speed over a shorter interval to prevent the robot falling off the ramp.Source: §5.2.2 · p. 14

5.2.3. Results and discussion

  • 2 m/s forward and 1.5 m/s laterally, both without slippage; lateral friction and vibration limited performance relative to forward motion.On soft terrain, maximum lateral velocity was 0.1 m/s; the robot also climbed slopes up to 20° at 0.1 m/s and descended at -0.1 m/s.Source: §5.2.3 · pp. 14–16
  • Forward-motion torque peaked during acceleration, while lateral torque remained high at constant velocity; soft terrain produced the highest maximum torque, but velocity stayed stable across forward and backward motion.Torque magnitude closely matched the estimate for forward motion at 2 m/s; grass torque fluctuated irregularly, likely due to uneven ground.Source: §5.2.3 · pp. 14–15, Figure 14 · p. 15, Figure 15 · p. 15

5.3. Overcome obstacle: going below

OmniQuad passed beneath a horizontal rod by lowering its base, moving laterally, and returning to its initial pose. It also climbed a 100 mm step using its legs and wheels, with leg trajectory errors below 1.5° throughout the task.

5.3.1. Experimental setup

  • The underpass setup used a rod longer than the robot, supported at both ends by boxes at a predetermined height of 200 mm.Source: §5.3.1 · p. 16

5.3.2. Test description

  • Under a 200 mm-high rod, the robot lowered its base to 180 mm, moved laterally beneath it, and raised its base back to its initial pose.The robot’s height differed from its reference by 20 mm in the shorter phase, while maximum joint-trajectory error was around 3°; lateral wheel rotation was in opposite directions.Source: §5.3.2 · p. 16, §5.3.1 · p. 16, §5.3.3 · p. 16

5.3.3. Results and discussion

  • During the underpass, lateral motion was smooth at lower speed; vibrations became perceptible as velocity increased but did not affect operation.The base-height tracking error in the shorter phase was 20 mm, attributed to the robot’s weight.Source: §5.3.3 · p. 16

5.4.1. Experimental setup

  • The step-climbing test used cushioned rigid boxes to create a uniform step surface and reduce terrain irregularities.Mats were placed on the boxes and nearby area to help keep the robot safe from sudden failure.Source: §5.4.1 · p. 17

5.4.2. Test description

  • The robot ascended a 100 mm step by placing its front legs over it, using its wheels to bring the hind legs to the step, and then readjusting its base orientation.It completed the final part using its wheels.Source: §5.4.2 · p. 17, Figure 20 · p. 18

5.4.3. Results and discussion

  • Less than 1.5° error separated reference and measured leg trajectories throughout the step-climbing task.The robot tilted its base to keep its center of gravity within the support polygon while lifting a foot with only three legs in contact.Source: §5.4.3 · pp. 17, 19, Figure 20 · p. 18

5.5. Joystick Experiments

In an outdoor grass trial, OmniQuad completed joystick-commanded forward movement, turns, and base-height changes on irregular terrain despite increased vibrations.

5.5.1. Experimental setup

  • The joystick controller was tested on grass, described as a natural surface frequently encountered in human environments.Source: §5.5.1 · p. 19

5.5.2. Test description

  • The joystick trial used a 3 m × 3 m operating grid and included forward movement, lowering the base, two 90° turns, and raising the base.The robot began at approximately 450 mm height, moved forward at 0.5 m/s, and reached a maximum angular velocity of 1.0 rad/s.Source: §5.5.2 · p. 19

5.5.3. Results and discussion

  • The robot completed the joystick task on irregular grass despite increased vibrations, indicating potential for effective outdoor performance.Source: §5.5.3 · p. 19

5.6. Obstacle detection experiment

OmniQuad used increased wheel torque on contact to trigger obstacle detection and a three-wheel traversal maneuver. It crossed the obstacle while maintaining locomotion with three wheels in contact with the ground.

5.6.1. Experimental setup

  • The obstacle was a 12 cm-high box measuring 250 mm × 150 mm, and the test assessed detection, posture adjustment, and controlled ascent during lateral motion.Source: §5.6.1 · pp. 19–20

5.6.2. Test description

  • Initial contact increased wheel torque and triggered obstacle detection; after the wheel torque reached zero, the robot lifted one wheel, moved diagonally on three wheels, then restored its nominal configuration and resumed lateral motion.Source: §5.6.2 · p. 20

5.6.3. Results and discussion

  • With one wheel lifted, OmniQuad maintained effective locomotion on the other three wheels during obstacle traversal.The front-left wheel’s torque peaked before it was lifted; the authors describe the result as demonstrating robustness of the mecanum wheels and actuation system.Source: §5.6.3 · p. 20, Figure 23 · p. 20, Figure 22 · p. 20

5.7. Stress Test

The stress tests examined obstacle traversal, durability, and payload capacity during legged locomotion. OmniQuad completed the 3.5 cm step with 2.5 kg, while a 3 kg attempt ended in driver overheating and a fall.

5.7.1. Experimental setup

  • The experiments assessed hardware and controller robustness, including durability and maximum legged-locomotion payload on a single 3.5 cm step.Payload increased from 0.5 kg to 2.5 kg in increments after successful trials, with obstacle detection and climbing enabled.Source: §5.7.1 · p. 21, §5.7.2 · p. 21

5.7.2. Test description

  • The durability course tested obstacle and surface adaptation through a transparent panel, a 5° slope, a 15 cm step-down, a foam mattress, and obstacle pushing.The robot was joystick-controlled and sometimes descended the slope laterally rather than following the intended path.Source: §5.7.2 · p. 21

5.7.3. Results and discussion

  • Two continuous durability runs lasting 5 and 6 minutes ended without hardware malfunctions or software failures.The joystick interface remained responsive during both trials.Source: §5.7.3 · p. 22
  • 2.5 kg was the maximum successful payload for the 3.5 cm step; at 3 kg, motor-driver overheating caused loss of control and a fall.During most of the successful step, joint torque remained below 10 Nm, against a 19.6 Nm maximum peak torque for the EM-Act actuators.Source: §5.7.3 · p. 22, Figure 25 · p. 22

5.8. Energy Consumption

Energy estimates varied with locomotion mode: wheels-only travel had the longest estimated battery life, while leg motion and hybrid obstacle crossing shortened it. Leg motors drew substantially more current than wheel motors in the reported calculation.

5.8. Energy Consumption

  • The energy estimate converts mean current into watt-hours using supply voltage and operation time.The calculation assumes a 100-second interval and 24 V supply; the passage gives a 99.90 Wh battery capacity.Source: §5.8 · pp. 22–23
  • Mean current was approximately 6.05 A per leg motor versus 0.44 A per wheel motor; together, eight leg and four wheel motors drew 50.2 A.At 24 V over 100 seconds, the estimated consumption was approximately 34.9 Wh.Source: §5.8 · p. 23
  • Estimated battery life was approximately 3 hours on wheels, 2 hours with base-lowering and -raising leg motion, and 1.5 hours for hybrid obstacle crossing.The base-motion case increased leg-motor current absorption by 50%.Source: §5.8 · pp. 22–23

5.9. Comparison with state-of-the-art approaches

Prior wheel-leg control work has variously emphasized mechanical design, leg design, or model-predictive control. OmniQuad’s approach pairs a compact, powerful wheeled foot with control intended to use wheels and legs for climbing and manoeuvrability.

5.9. Comparison with state-of-the-art approaches

  • Prior wheel-leg composite-control approaches target stability, efficiency, and terrain adaptability: prioritizes mechanical design and simplified omnidirectional wheeled locomotion with minimal leg movement, while targets impact mitigation and balance on uneven terrains.[34] concentrates on leg design to minimize impact.Source: §5.9 · p. 23
  • Other cited work uses MPC for travel-time optimization under car-like steering constraints or whole-body stability control for a six-legged robot.These approaches are described in and, respectively.Source: §5.9 · p. 24
  • OmniQuad’s design focus is a compact, powerful wheeled foot, paired with control that uses wheels and legs for climbing and manoeuvrability.The stated aim is navigation in challenging environments.Source: §5.9 · p. 24

6. Conclusion and Future Works

OmniQuad combines mecanum wheels and legs for omnidirectional movement, obstacle traversal, and height adjustment. The authors report current autonomy in structured settings with static obstacles and identify design and navigation capabilities for future work.

6. Conclusion and Future Works

  • OmniQuad combines mecanum wheels and legs, reducing its height by up to 180 mm and ascending steps as high as 100 mm under varying payload conditions.The conclusion also reports precise movement while maintaining orientation and directionality.Source: §6 · p. 24
  • The robot was presented at Maker Faire Rome 2023 and the European Robotics Forum 2024 in Rimini, Italy.Source: §6 · p. 24
  • The authors identify structured environments with static obstacles as the scope of current encoder-based odometry and orientation-based autonomy.They describe this autonomy as sufficient for simple scenarios, while targeting unknown and dynamic environments with depth-camera and LiDAR sensing.Source: §6 · p. 24
  • Future work includes a shorter wheeled foot, hip-abduction joints, and improved navigation for unknown or dynamic environments.Planned work also addresses terrain recognition, dynamic balance, and efficient transitions between locomotion modes.Source: §6 · pp. 24–25
Loading 2610.38502v1…