Skip to main content

Table 2 Overview of lower limb EMG-driven controllers working principles

From: EMG-driven control in lower limb prostheses: a topic-based systematic review

Ref. Walking controller Slope/speed adaptation Additional modalities Training/calibration time
Direct control
[64] EMG-triggered knee joint lock during stance phase SlA\(^{\star }\), SpA\(^{\star }\) All (STA\(^{\star }\)) NN
[29] EMG-proportional modulation of knee joint velocity SlA, SpA\(^{\star }\) All (not tested) NS
[24,25,26,27] ML-driven knee joint angle trajectory generation SlA, SpA All (not tested) CT: 10–15 s, per 2 sessions, per 5 days
[56] EMG-driven knee joint stiffness set-point SlA, SpA All (NWB\(^{\star }\)) ST: 1 h, before each use
[14, 23, 62, 63] EMG-driven knee joint stiffness set-point SlA, SpA All (STND\(^{\star }\), SIT\(^{\star }\), SQ\(^{\star }\), STA\(^{\star }\), NWB\(^{\star }\)) ST: 3 h, per 4 sessions; CT: 2 h trajectory tracking trials
[57] EMG-driven multi-DoF knee and ankle joint stiffness set-point SlA, SpA All (NWB\(^{\star }\)) ST: therapist session; CT: 3 s per 64 trials, per 4 sessions
[37] ML-driven knee joint angle trajectory generation SlA, SpA All (not tested) NS
[16] EMG-driven ankle joint stiffness set-point SlA, SpA All (NWB\(^{\star }\)) CT: 10 trials (\(\sim\) 80 s)
[67, 68] EMG-proportional plantarflexor torque generation SlA, SpA All (not tested) CT: NS
[76] EMG-triggered ankle plantarflexion and dorsiflexion NI NI CT: NS
[11] EMG-decoded ankle and knee joint angle trajectory generation SlA\(^{\star }\) All (STA\(^{\star }\), STD\(^{\star }\)) CT: \(\sim\) 20 trials per task
[39, 40] EMG-proportional plantarflexor torque generation SlA, SpA All ST: limited acclimation period
Pattern recognition control
[98] EMG-driven knee FSM (Stance [Post-HS, FF and Pre-TO], swing [SF, SE]) SlA NI Adaptation period of 20 min; FSM CT: NS
[70] Knee joint moment control as function of EMG-driven locomotion identification SlA\(^{\star }\), SpA\(^{\star }\) STA\(^{\star }\), STD\(^{\star }\) FSM CT: NS
[5] EMG-driven FSM for level ground walking and stairs climbing SlA, SpA STA\(^{\star }\) FSM CT: NS; ST < 20 min
[65, 66, 141, 144] ML-driven knee joint FSM (Stance [Post-HS, Pre-TO], swing [Post-TO, Pre-HS]) SlA\(^{\star }\) OBST\(^{\star }\), STND\(^{\star }\), STA\(^{\star }\), STD\(^{\star }\), TURN\(^{\star }\) FSM CT: \(\sim\) 15 min (3 times each task)
[53] CPG-generated knee and ankle joint trajectories as function of ML-driven locomotion identification NI STND\(^{\star }\), SIT\(^{\star }\), STA\(^{\star }\), STD\(^{\star }\) FSM CT: NS
[85] ML-driven knee joint FSM (Stance [Post-HS, FF, Pre-TO], swing [Post-TO, Pre-HS]) NI STA\(^{\star }\), STD\(^{\star }\) FSM CT: 50 gait cycles per task
[15, 110] ML-driven knee joint FSM (Stance [Post-HS, FF, Pre-TO], swing [Post-TO, Pre-HS]) NI NI FSM CT: 70 gait cycles
[32, 33, 143] ML-driven knee joint FSM (Stance [Post-HS, Pre-TO], swing [Post-TO, Pre-HS]) SlA\(^{\star }\) STA\(^{\star }\), STD\(^{\star }\) ST: therapist sessions; FSM CT: \(\sim\) 30s (5 times per task)
[86, 87] ML-driven ankle joint FSM (Stance [Post-HS, Pre-TO], swing) SlA\(^{\star }\), SpA\(^{\star }\) STA\(^{\star }\), STD\(^{\star }\) FSM CT: 21 trials in total, 6–7 steps per trial
[120] ML-driven FSM for multi-DoF ankle joint SlA, SpA All (NWB\(^{\star }\)) FSM CT: 3 s per 8 trial, per 7 tasks
[58, 111] ML-driven knee joint FSM (Stance [Post-HS, Pre-TO], swing [Post-TO, Pre-HS]) NI STND\(^{\star }\), SIT\(^{\star }\), NWB\(^{\star }\) FSM CT: NS
[59, 113, 114, 139, 140] Knee and ankle joint impedance characterization as function of ML-driven locomotion identification SlA\(^{\star }\) STA\(^{\star }\), STD\(^{\star }\), SIT\(^{\star }\), NWB\(^{\star }\) Intrinsic controller parameters tuning (NS); FSM CT: 10–20 trials per task
[17] ML-driven ankle joint impedance characterization based terrain inclination classification SlA\(^{\star }\) NI Intrinsic controller parameters tuning (NS); CT: 3 sessions; ST: \(\sim\) 5 h
[106] EMG-triggered knee joint motion routine NI NI NS
[54, 55] ML-driven ankle joint FSM SlA\(^{\star }\) STA\(^{\star }\), STD\(^{\star }\) FSM CT: 5 gait cycles per trial; ST: 5 min per task
Model-based control
[6] EMG-driven model-based ankle joint angle trajectory generation SlA, SpA All (NWB\(^{\star }\)) Virtual environment training: NS
[61] EMG-driven model-based knee joint impedance characterization SlA, SpA All (not tested) CT: NS
[135, 136] EMG-driven model-based knee joint impedance characterization SlA, SpA All (NWB\(^{\star }\)) CT: trajectory tracking trials, walking experiments
[71, 72, 130] EMG-modulation of model-based ankle joint moment trajectory SlA, SpA All (STA\(^{\star }\), STD\(^{\star }\)) CT: 10 steps
[19] Hybrid ML-NMS model-based knee joint moment generation SlA, SpA\(^{\star }\) All(STND\(^{\star }\), SIT\(^{\star }\)) CT: 3–10 trials per motor task
  1. Fields include: paper reference; walking controller (the high-level control law during the walking cycle); slope/speed adaptation (the ability of the walking controller to adapt to different slope angles and ambulation velocities); additional modalities (additional types of locomotion supported from the EMG-driven controller); training/calibration time (required time to either calibrate the controller parameters or train the subject)
  2. NN not necessary, NS not stated, NI not implemented, ML machine learning, NMS neuromuscularskeletal, CPG central pattern generator, HS heel strike, FF foot flat, TO toe off, SF swing flexion, SE swing extension, FSM finite-state machine, DoF Degrees of freedom, SlA slope adaptation, SpA speed adaptation, All no restriction in the locomotion control, NWB non-weight bearing joint movements joint movement, STND standing, SIT sitting, SQ squatting, STA stairs ascending, STD stairs descending, OBST obstacle stepping, TURN turning on the spot, CT calibration time, ST subject training
  3. \(^{\star }\)Tested modalities