Lower limb exoskeleton systems-overview¹

Assistive

Intention Estimation ​

1. Direct

   • estimation from biological signals
     • biological signals are often noisy and difficult to measure
     • more accessible signals such as EMG not usable for the lost connection between the brain and the limbs.

2. Indirect

   • the center of mass (COM) velocity, posture(the torso angle; the center of pressure (COP)), etc.

     • Additional sensors such as IMU or GRF sensors

       no need of high sample frequency or filtering since they only provide a rough estimation of human intention.

Trajectories generating

Generated trajectories may not be identical to a healthy subject’s motion profile, but they can still complete the task, e.g. unactuated or fixed ankle joints.

• a finite state machine

  dividing the single full motion cycle into different phases. For each phase, the controller may have a different control scheme. ($\Leftarrow$ different phases of walking present different dynamics.)

further research

• sit-to-stand function
• flexible control(adding a compliant assisting mode to partially preserve users’ walking ability; variable stiffness actuator to further control the torque and stiffness at each joint)
• online calculation of motion profile
• relationship between different joints
• balance
  • extra support (support besides legs);
  • estimating the XCoM (the position of the center of mass combined with momentum, during walking)
  • properties of human limbs, such as center of mass, inertia, and link length need to be precisely measured

Rehabilitation

with weight support (on treadmills/body weight supported with overhead cables/harness)

• constraints to assure correct gait patterns

• orthoses to prevent possible injuries caused by incorrect leg configuration since lower limbs are redundant in the sagittal plane.

• compliance: rehabilitation exoskeletons focus on human-exoskeleton contact and interaction with the user.

  • simulating with control, i.e. impedance or admittance control (input/output relationship is position/force or force/position respectively)
    • In impedance control, a common scheme is to create a virtual environment and determine torque output by simulating the robot as interacting with a compliant material
  • adding an elastic component in the actuators

• “assist as needed”

Safety

• having the user use a walker

• high-power density actuators for user weight supporting

  Me: 为什么气动属于high-power density？这个怎么判断

  power density 是功率密度，输出功率/质量

Control

• mostly through impedance control

Further

• functional electrical stimulation (FES) to cause the user to increase their effort $\Rightarrow$ “hybrid exoskeletons”

Augmentation

Assistance directly applied to tasks(that a normal human cannot)

e.g. for loading/unloading the missiles of fighter jets on an aircraft carrier, or carrying extra load while hiking

Disadvantages: often significantly increases the overall weight and reduces the range of motion of the human-exoskeleton system.

Control

• Sensitivity amplification (the human and exoskeleton as one system)
  • short rise time and high overshoot
  • eliminating the problem of donning and misplacement of the sensors
  • balance problems if sensitivity increased
• Position control
  • help to stabilize the system
• EMG
  • only category of sensing that can estimate the user’s intention without a time delay
  • limited efforts of EMG-based systems $\Leftarrow$ the more assistance the user receives, the less bioelectrical signals he/she will generate
• combining active and passive systems

Assistance applied to users

Pros & cons: simple, light, and more comfortable; tasks will be limited by users’ physical capability since it will first interact with the user.

Goal: to properly assist humans so that they can perform the same task with a lower metabolic cost

• energy efficiency
  • bipedal walking: Other than for overcoming friction, passive walkers do not need additional effort to walk
  • humans: significant energy loss in the step-to-step transition;
    • The common explanation is smoothly switching the center of mass between steps is difficult for humans because the vertical projection of the COM (known as COP) has to move out of the supporting plane (feet), resulting in intentional falling and creating impact when the other leg touches the ground. This induces energy lost during the transition state and is often referred to as “heel strike,” which is the primary source of energy lost during ground level walking.

Researches

• attempts to replace muscle activity with the exoskeleton’s assistance

  • challenging to provide correct assistance
  • assistance from the exoskeleton may cause unknown energy consumption if it is unnatural²³

• oscillator-based control require periodic motion(not fit for some tasks like squatting)

• limited efficiency of systems driven by electric motors(most are servo motors) $\Leftarrow$ fundamental difference between motors and human joints

  For example, the human knee joint angular position may range from 120 to 1110 degrees, and the angular velocity generally is less than 30 rpm but with high torque output. Alternatively, most motors operate in 360 degrees and have maximized efficiency and torque output at a specific rpm.

  • gear reduction & dilemma: a high gear ratio creates larger torque but reduces angular velocity in human joints.
  • motor efficiency Alò et al.⁴ designed a transmission system containing a fly wheel with an infinitely variable transmission between the motors and the knee joint; yet the effect of the extra weight on the system is unknown.

• harvest energy from human motion

  • Malcolm et al.⁵: applying assistive torque at a specific time (i.e., at the supporting leg, just before the heel strike of the swinging leg) without using energy recycled from the negative work inherent in bipedal walking

  • orthosis design⁶

  • Collins et al.: a passive device using a clutch and spring system to assist the ankle joint⁷

    ankle joints are the most energy-efficient joint, as they have the Achilles tendon to store negative work. Thus, the potential for improvements is not as high as for other joints.

  • knee: Rogers et al.⁸, a quasi-passive air spring to absorb the impulse; Yamada et al.⁹, a pneumatic walking assistive device that used the energy recycled from the knee joint during the beginning of the double support phase and reused it on the same joint later.

  • optimal design: to be apply energy recycled from the joints that produce the most negative work on those that require the most positive work

    • CYBERLEGs Alpha-Prototype¹⁰ transfer the energy from knee to ankle; van den Bogert¹¹ and Van Dijk¹² connected multiple joints of their exoskeletons with artificial tendons.
    • However, the metabolic costs were not reduced due to extra restrictions that affected human motion.

Actuation

Future for lower limb exoskeletons

1. Operating time
   • Larger energy sources (e.g., batteries) with higher mass not ideal
   • Energy harvesting and recycling systems
2. Human-exoskeleton interaction
   • dilemma: exoskeletons' motion profiles need to match the user (delay) vs interaction needed to “feel” the user’s intention (instability)
   • To reduce undesired interaction (additional mass or less range of motion) $\Rightarrow$ uncomfortable & larger metabolic cost
     • current improvements are mainly focusing on introducing compliance in the actuators (soft exoskeleton) and reducing weight yet of which the controllability and support are limited.

Appendix

Some Assistive Exoskeletons

Some Rehabilitation Exoskeletons

Some Augmentation Exoskeletons

References