Optimal feedback control successfully explains changes in neural modulations during experiments with brain-machine interfaces.
ABSTRACT: Recent experiments with brain-machine-interfaces (BMIs) indicate that the extent of neural modulations increased abruptly upon starting to operate the interface, and especially after the monkey stopped moving its hand. In contrast, neural modulations that are correlated with the kinematics of the movement remained relatively unchanged. Here we demonstrate that similar changes are produced by simulated neurons that encode the relevant signals generated by an optimal feedback controller during simulated BMI experiments. The optimal feedback controller relies on state estimation that integrates both visual and proprioceptive feedback with prior estimations from an internal model. The processing required for optimal state estimation and control were conducted in the state-space, and neural recording was simulated by modeling two populations of neurons that encode either only the estimated state or also the control signal. Spike counts were generated as realizations of doubly stochastic Poisson processes with linear tuning curves. The model successfully reconstructs the main features of the kinematics and neural activity during regular reaching movements. Most importantly, the activity of the simulated neurons successfully reproduces the observed changes in neural modulations upon switching to brain control. Further theoretical analysis and simulations indicate that increasing the process noise during normal reaching movement results in similar changes in neural modulations. Thus, we conclude that the observed changes in neural modulations during BMI experiments can be attributed to increasing process noise associated with the imperfect BMI filter, and, more directly, to the resulting increase in the variance of the encoded signals associated with state estimation and the required control signal.
Project description:This paper presents a novel inverse kinematics solution for robotic arm based on artificial neural network (ANN) architecture. The motion of robotic arm is controlled by the kinematics of ANN. A new artificial neural network approach for inverse kinematics is proposed. The novelty of the proposed ANN is the inclusion of the feedback of current joint angles configuration of robotic arm as well as the desired position and orientation in the input pattern of neural network, while the traditional ANN has only the desired position and orientation of the end effector in the input pattern of neural network. In this paper, a six DOF Denso robotic arm with a gripper is controlled by ANN. The comprehensive experimental results proved the applicability and the efficiency of the proposed approach in robotic motion control. The inclusion of current configuration of joint angles in ANN significantly increased the accuracy of ANN estimation of the joint angles output. The new controller design has advantages over the existing techniques for minimizing the position error in unconventional tasks and increasing the accuracy of ANN in estimation of robot's joint angles.
Project description:Real-time brain-machine interfaces (BMI) have focused on either estimating the continuous movement trajectory or target intent. However, natural movement often incorporates both. Additionally, BMIs can be modeled as a feedback control system in which the subject modulates the neural activity to move the prosthetic device towards a desired target while receiving real-time sensory feedback of the state of the movement. We develop a novel real-time BMI using an optimal feedback control design that jointly estimates the movement target and trajectory of monkeys in two stages. First, the target is decoded from neural spiking activity before movement initiation. Second, the trajectory is decoded by combining the decoded target with the peri-movement spiking activity using an optimal feedback control design. This design exploits a recursive Bayesian decoder that uses an optimal feedback control model of the sensorimotor system to take into account the intended target location and the sensory feedback in its trajectory estimation from spiking activity. The real-time BMI processes the spiking activity directly using point process modeling. We implement the BMI in experiments consisting of an instructed-delay center-out task in which monkeys are presented with a target location on the screen during a delay period and then have to move a cursor to it without touching the incorrect targets. We show that the two-stage BMI performs more accurately than either stage alone. Correct target prediction can compensate for inaccurate trajectory estimation and vice versa. The optimal feedback control design also results in trajectories that are smoother and have lower estimation error. The two-stage decoder also performs better than linear regression approaches in offline cross-validation analyses. Our results demonstrate the advantage of a BMI design that jointly estimates the target and trajectory of movement and more closely mimics the sensorimotor control system.
Project description:Stability of human gait is the ability to maintain upright posture during walking against external perturbations. It is a complex process determined by a number of cross-related factors, including gait trajectory, joint impedance and neural control strategies. Here, we consider a control strategy that can achieve stable steady-state periodic gait while maintaining joint flexibility with the lowest possible joint impedance. To this end, we carried out a simulation study of a heel-toe footed biped model with hip, knee and ankle joints and a heavy head-arms-trunk element, working in the sagittal plane. For simplicity, the model assumes a periodic desired joint angle trajectory and joint torques generated by a set of feed-forward and proportional-derivative feedback controllers, whereby the joint impedance is parametrized by the feedback gains. We could show that a desired steady-state gait accompanied by the desired joint angle trajectory can be established as a stable limit cycle (LC) for the feedback controller with an appropriate set of large feedback gains. Moreover, as the feedback gains are decreased for lowering the joint stiffness, stability of the LC is lost only in a few dimensions, while leaving the remaining large number of dimensions quite stable: this means that the LC becomes saddle-type, with a low-dimensional unstable manifold and a high-dimensional stable manifold. Remarkably, the unstable manifold remains of low dimensionality even when the feedback gains are decreased far below the instability point. We then developed an intermittent neural feedback controller that is activated only for short periods of time at an optimal phase of each gait stride. We characterized the robustness of this design by showing that it can better stabilize the unstable LC with small feedback gains, leading to a flexible gait, and in particular we demonstrated that such an intermittent controller performs better if it drives the state point to the stable manifold, rather than directly to the LC. The proposed intermittent control strategy might have a high affinity for the inverted pendulum analogy of biped gait, providing a dynamic view of how the step-to-step transition from one pendular stance to the next can be achieved stably in a robust manner by a well-timed neural intervention that exploits the stable modes embedded in the unstable dynamics.
Project description:Movement planning is thought to be primarily determined by motor costs such as inaccuracy and effort. Solving for the optimal plan that minimizes these costs typically leads to specifying a time-varying feedback controller which both generates the movement and can optimally correct for errors that arise within a movement. However, the quality of the sensory feedback during a movement can depend substantially on the generated movement. We show that by incorporating such state-dependent sensory feedback, the optimal solution incorporates active sensing and is no longer a pure feedback process but includes a significant feedforward component. To examine whether people take into account such state-dependency in sensory feedback we asked people to make movements in which we controlled the reliability of sensory feedback. We made the visibility of the hand state-dependent, such that the visibility was proportional to the component of hand velocity in a particular direction. Subjects gradually adapted to such a sensory perturbation by making curved hand movements. In particular, they appeared to control the late visibility of the movement matching predictions of the optimal controller with state-dependent sensory noise. Our results show that trajectory planning is not only sensitive to motor costs but takes sensory costs into account and argues for optimal control of movement in which feedforward commands can play a significant role.
Project description:Animals vary considerably in size, shape, and physiological features across individuals, but yet achieve remarkably similar behavioral performances. We examined how animals compensate for morphophysiological variation by measuring the system dynamics of individual knifefish (Eigenmannia virescens) in a refuge tracking task. Kinematic measurements of Eigenmannia were used to generate individualized estimates of each fish's locomotor plant and controller, revealing substantial variability between fish. To test the impact of this variability on behavioral performance, these models were used to perform simulated 'brain transplants'-computationally swapping controllers and plants between individuals. We found that simulated closed-loop performance was robust to mismatch between plant and controller. This suggests that animals rely on feedback rather than precisely tuned neural controllers to compensate for morphophysiological variability.
Project description:During planning and execution of reaching movements, the activity of cortical motor neurons is modulated by a diversity of motor, sensory, and cognitive signals. Brain-machine interfaces (BMIs) extract part of these modulations to directly control artificial actuators. However, cortical modulations that emerge in the novel context of operating the BMI are poorly understood.Here we analyzed the changes in neuronal modulations that occurred in different cortical motor areas as monkeys learned to use a BMI to control reaching movements. Using spike-train analysis methods we demonstrate that the modulations of the firing-rates of cortical neurons increased abruptly after the monkeys started operating the BMI. Regression analysis revealed that these enhanced modulations were not correlated with the kinematics of the movement. The initial enhancement in firing rate modulations declined gradually with subsequent training in parallel with the improvement in behavioral performance.We conclude that the enhanced modulations are related to computational tasks that are significant especially in novel motor contexts. Although the function and neuronal mechanism of the enhanced cortical modulations are open for further inquiries, we discuss their potential role in processing execution errors and representing corrective or explorative activity. These representations are expected to contribute to the formation of internal models of the external actuator and their decoding may facilitate BMI improvement.
Project description:Control of our movements is apparently facilitated by an adaptive internal model in the cerebellum. It was long thought that this internal model implemented an adaptive inverse model and generated motor commands, but recently many reject that idea in favor of a forward model hypothesis. In theory, the forward model predicts upcoming state during reaching movements so the motor cortex can generate appropriate motor commands. Recent computational models of this process rely on the optimal feedback control (OFC) framework of control theory. OFC is a powerful tool for describing motor control, it does not describe adaptation. Some assume that adaptation of the forward model alone could explain motor adaptation, but this is widely understood to be overly simplistic. However, an adaptive optimal controller is difficult to implement. A reasonable alternative is to allow forward model adaptation to 're-tune' the controller. Our simulations show that, as expected, forward model adaptation alone does not produce optimal trajectories during reaching movements perturbed by force fields. However, they also show that re-optimizing the controller from the forward model can be sub-optimal. This is because, in a system with state correlations or redundancies, accurate prediction requires different information than optimal control. We find that adding noise to the movements that matches noise found in human data is enough to overcome this problem. However, since the state space for control of real movements is far more complex than in our simple simulations, the effects of correlations on re-adaptation of the controller from the forward model cannot be overlooked.
Project description:Functional electrical stimulation (FES), the coordinated electrical activation of multiple muscles, has been used to restore arm and hand function in people with paralysis. User interfaces for such systems typically derive commands from mechanically unrelated parts of the body with retained volitional control, and are unnatural and unable to simultaneously command the various joints of the arm. Neural interface systems, based on spiking intracortical signals recorded from the arm area of motor cortex, have shown the ability to control computer cursors, robotic arms and individual muscles in intact non-human primates. Such neural interface systems may thus offer a more natural source of commands for restoring dexterous movements via FES. However, the ability to use decoded neural signals to control the complex mechanical dynamics of a reanimated human limb, rather than the kinematics of a computer mouse, has not been demonstrated. This study demonstrates the ability of an individual with long-standing tetraplegia to use cortical neuron recordings to command the real-time movements of a simulated dynamic arm. This virtual arm replicates the dynamics associated with arm mass and muscle contractile properties, as well as those of an FES feedback controller that converts user commands into the required muscle activation patterns. An individual with long-standing tetraplegia was thus able to control a virtual, two-joint, dynamic arm in real time using commands derived from an existing human intracortical interface technology. These results show the feasibility of combining such an intracortical interface with existing FES systems to provide a high-performance, natural system for restoring arm and hand function in individuals with extensive paralysis.
Project description:Recent studies show that the amplitude of cortical field potentials is modulated in the time domain by grasping kinematics. However, it is unknown if these low frequency modulations persist and contain enough information to decode grasp kinematics in macro-scale activity measured at the scalp via electroencephalography (EEG). Further, it is unclear as to whether joint angle velocities or movement synergies are the optimal kinematics spaces to decode. In this offline decoding study, we infer from human EEG, hand joint angular velocities as well as synergistic trajectories as subjects perform natural reach-to-grasp movements. Decoding accuracy, measured as the correlation coefficient (r) between the predicted and actual movement kinematics, was r = 0.49 ± 0.02 across 15 hand joints. Across the first three kinematic synergies, decoding accuracies were r = 0.59 ± 0.04, 0.47 ± 0.06, and 0.32 ± 0.05. The spatial-temporal pattern of EEG channel recruitment showed early involvement of contralateral frontal-central scalp areas followed by later activation of central electrodes over primary sensorimotor cortical areas. Information content in EEG about the grasp type peaked at 250 ms after movement onset. The high decoding accuracies in this study are significant not only as evidence for time-domain modulation in macro-scale brain activity, but for the field of brain-machine interfaces as well. Our decoding strategy, which harnesses the neural "symphony" as opposed to local members of the neural ensemble (as in intracranial approaches), may provide a means of extracting information about motor intent for grasping without the need for penetrating electrodes and suggests that it may be soon possible to develop non-invasive neural interfaces for the control of prosthetic limbs.
Project description:BACKGROUND: Spasticity is an important complication after stroke, especially in the anti-gravity muscles, i.e. lower limb extensors. However the contribution of hyperexcitable muscle spindle reflex loops to gait impairments after stroke is often disputed. In this study a neuro-musculoskeletal model was developed to investigate the contribution of an increased length and velocity feedback and altered reflex modulation patterns to hemiparetic gait deficits. METHODS: A musculoskeletal model was extended with a muscle spindle model providing real-time length and velocity feedback of gastrocnemius, soleus, vasti and rectus femoris during a forward dynamic simulation (neural control model). By using a healthy subject's base muscle excitations, in combination with increased feedback gains and altered reflex modulation patterns, the effect on kinematics was simulated. A foot-ground contact model was added to account for the interaction effect between the changed kinematics and the ground. The qualitative effect i.e. the directional effect and the specific gait phases where the effect is present, on the joint kinematics was then compared with hemiparetic gait deviations reported in the literature. RESULTS: Our results show that increased feedback in combination with altered reflex modulation patterns of soleus, vasti and rectus femoris muscle can contribute to excessive ankle plantarflexion/inadequate dorsiflexion, knee hyperextension/inadequate flexion and increased hip extension/inadequate flexion during dedicated gait cycle phases. Increased feedback of gastrocnemius can also contribute to excessive plantarflexion/inadequate dorsiflexion, however in combination with excessive knee and hip flexion. Increased length/velocity feedback can therefore contribute to two types of gait deviations, which are both in accordance with previously reported gait deviations in hemiparetic patients. Furthermore altered modulation patterns, in particular the reduced suppression of the muscle spindle feedback during swing, can contribute largely to an increased plantarflexion and knee extension during the swing phase and consequently to hampered toe clearance. CONCLUSIONS: Our results support the idea that hyperexcitability of length and velocity feedback pathways, especially in combination with altered reflex modulation patterns, can contribute to deviations in hemiparetic gait. Surprisingly, our results showed only subtle temporal differences between length and velocity feedback. Therefore, we cannot attribute the effects seen in kinematics to one specific type of feedback.