United States

In this work we present a spiking neural network for the adaptive control of a simulated Kinova Gen3 7-DOF arm. This work extends the Recurrent Error-driven Adaptive Control Hierarchy (REACH), developed as a computational model of the primate motor control system, which had previously been implemented for the operational space control of a 3-DOF arm restricted to the horizontal plane. The REACH model is divided into kinematics control, responsible for transforming the control signal from end-effector (hand) space to joint-space, and dynamics control, responsible for accounting for the effects of gravity and inertia on the system. Additionally, a nonlinear adaptive control component learns online to correct for errors that arise during movement as a result of internal model inaccuracies or external perturbations. Due to the increased complexity of the 7-DOF system, the process for training the REACH model required a different approach than the 3-DOF implementation. Here, we utilize a hybrid approach to training that takes advantage of both deep learning and the circuit building methods from the Neural Engineering Framework (NEF). The NEF allows us to both quickly build highly optimized components performing computations on low-dimensional input, as well as integrate these networks directly with those trained using deep learning, which provides a more exhaustive sampling of high-dimensional state space. We analyze the performance of the spiking control network on a benchmark task compared to both an analytic controller and a rate-mode version of the control network. Lastly, we discuss the process of training the network to run on the highly-efficient neuromorphic chip from Intel, the Loihi, and provide initial results from simulation in a Loihi emulator.

NAML 2022

Adaptive spiking control of a 7 DOF arm

neurorobotic

adaptive control

spiking neural network

Naval Information Warfare Center Pacific is partnering with AFCEA San Diego chapter to host the workshop.

The NAML workshop is an opportunity for researchers from Navy research and development centers and our partner organizations to exchange technical ideas and to come together as a team.

The goals of NAML are:

* To spread awareness of current machine learning research relevant to Naval applications

* To connect machine learning researchers with experts in military requirements

* To build and strengthen collaborations within the defense research community

**Watch this video to hear about the AFCEA purpose.**

<iframe width=1280 height=706 frameborder="0" scrolling="no" src="https://screencast-o-matic.com/player/creleVVcvRh?v=6&ff=1&title=0&controls=1" allowfullscreen="true"></iframe>

NAML 2021

technical paper

##### For full functionality of Underline, please SHUT OFF your VPN, or use a personal device.


Naval Information Warfare Center Pacific is partnering with AFCEA San Diego chapter to host the workshop.

The NAML workshop is an opportunity for researchers from Navy research and development centers and our partner organizations to exchange technical ideas and to come together as a team.

The goals of NAML are:

To spread awareness of current machine learning research relevant to Naval applications

To connect machine learning researchers with experts in military requirements

To build and strengthen collaborations within the defense research community


<figure class="video_container">
  <iframe src="https://s3.amazonaws.com/pf-user-files-01/u-59356/uploads/2022-03-22/hk13mxo/NAML_ReceptionWithtitle.mp4" width=826  height=526  frameborder="0" allowfullscreen="true"> </iframe>
</figure>

Naval Information Warfare Center Pacific is partnering with AFCEA San Diego chapter to host the workshop. The NAML workshop is an opportunity for researchers from Navy research and development centers and our partner organizations to exchange technical ideas and to come together as a team.

The goals of NAML are:

To spread awareness of current machine learning research relevant to Naval applications

To connect machine learning researchers with experts in military requirements

To build and strengthen collaborations within the defense research community



Mitigation of range-restricted communication blocks in a distributed swarm is of interest to the Navy, due to the existence of data sharing limitations in all types of sensors and missions carried out in communications-denied environments. If a distributed swarm is affected by a communications block or delay, individual members should still be able to cooperatively identify, evaluate, and act as a team to preserve mission success based on a number of critical parameters guiding the scenario. In our approach, individual members of a heterogeneous swarm in a communication blocked environment construct their own perceived Common Operating Picture (COP). We modified a previously proposed task allocation algorithm to account for incomplete member COPs, and implemented a communications-limiting algorithm in simulation space that emulates a real-world communications block. This block impacts the decision making and asset prioritization of the task allocation algorithm, and our work contributes to a better understanding of task allocation and swarm regrouping in a limited communications environment. Results suggest that compensating for a swarm’s incomplete COP in a limited communications environment achieves a higher likelihood of mission success, and that a quasi-centralized approach for task dissemination ultimately performed more reliably than a fully decentralized approach.

Performative Analysis of Sensor Limitations in Communications Denied Autonomous Swarms

Tactical mission planning for air combat missions is a human-intensive procedure to determine how to coordinate a diverse set of assets. This coordination becomes increasingly difficult due to the physical constraints of the system and the multiple time scales. This paper borrows from safety-critical control to formulate a mathematical description of a class of tactical mission plans. This mathematical description results in a Hamilton-Jacobi-Isaacs (HJI) partial differential equation (PDE). However, due to the curse of dimensionality, numerical solutions to this high dimensional PDE are computationally intensive. To produce a numerical solution, the HJI equation is solved using deep learning. This formulation illustrates the effectiveness and utility of differential games for mission planning through the solution of two games. The first game, the Air3D game, is a collision avoidance scenario between two vehicles. The second is the active target defense (ATD) game. An infinite time solution has been recently found in the controls literature. This paper determines a finite time approximation that is compared with the infinite time solution. In particular, the region of space that the defender protects in a finite time horizon is numerically estimated.

Misson Planning Using Differential Games and Deep Learning

Fast reorientation of space-based sensors is needed to increase ISR capacity and to respond to high-value targeting requests. Rapid reorientation maneuvers are realized by constructing and solving a nonlinear optimization problem that models the physics of the rotational motion as part of the ground process of mission operations. The solution trajectories are then uplinked to the spacecraft for attitude guidance. Attitude guidance commands for a spacecraft are a time-series of incremental rotations that are given to the attitude control system at the closed-loop feedback rate which is typically greater than 10 Hz.

In practical space systems, it is generally not possible to uplink reorientation commands as a finely sampled time-series (at or greater than 10 Hz) since this would quickly exhaust the capacity of a spacecraft’s command buffer. Instead, the current paradigm is to issue a reorientation command as a single time-tagged target orientation. Then, an on-board trajectory generator is used to construct an appropriate rotation command time-series at the closed-loop feedback rate which are then fed to the attitude guidance system. Typically, these on-board trajectory generators do not have the processing power to perform trajectory optimization and instead generate a feasible, but sub-optimal, time-series command set at the appropriate rate.

A potential solution to this problem is to compute and then compress the optimized time-series commands using an autoencoder. An autoencoder is an unsupervised machine learning application of a neural network that conducts a non-linear mapping of the inputs to a lower dimensional space, called the latent space, using the encoder portion of the autoencoder. It then reconstructs the same input from the latent space using the decoder portion of the autoencoder. With such a method, the optimized command sequence can be generated and encoded on the ground, then only the latent space is transmitted to the spacecraft and decoded on-board the spacecraft resulting in the reconstructed optimized rotation sequence at the closed-loop feedback rate. Given the time-series nature of the optimized command sequence, one appropriate architecture for the autoencoder is a long short term memory (LSTM) neural network autoencoder.

In this talk, we demonstrate the methodology for constructing and training the LSTM autoencoder and show the reconstruction error for different degrees of compression represented by decreasing dimensionality of the latent space. We then compare the reconstructed command sequence and the uncompressed time-optimal solutions using a flight simulation software at the closed-loop feedback rate. The result is a new and computationally efficient approach for implementing non-legacy maneuvers on board a spacecraft. There are numerous applications beyond spacecraft where this capability can benefit the DoD such as small unmanned vehicle operations and operations in degraded electromagnetic environments.

A Method to Compress and Transmit Optimal Reorientation Commands to a Spacecraft Using Long Short-Term Memory Neural Network Autoencoders

Current approaches for handling machinery control system (MCS) faults on unmanned surface vessels (USVs) are based on guidance and thresholds selected by subject matter experts. While this can be a robust method, it requires significant human input and validation testing in expected operating conditions. To provide vessels with greater autonomy, data driven approaches for MCS faults are of interest. Data driven approaches based on AI/ML, besides saving both time and money, may provide a constructive means to learn from past experience, learn new insights, and optimally handle faults based on the current environment and conditions (e.g. vessel speed, payload, or sea state). One significant challenge with using AI/ML approaches for fault handling is the issue of class balance – in general, one may have limited realizations or samples of a given fault. In order to positively utilize data from previous successful test events, this work proposes the casting the problem of fault detection as an outlier detection problem. By building a healthy state model, detection of deviations from the nominal state provide a means to recognize new or novel operating conditions or the failure or degradation of a given subsystem. By quantifying and tracking the healthy state, long-term trends and typical behavior can be identified. These metrics could also be used to schedule inspections and maintenance based on the condition of the vessel as opposed to predetermined ones. With most outlier detection methods, typically a given threshold probability needs to be selected. Optimal selection of this threshold requires observations of such an outlier. Given that many of the sensors used to trigger hull, mechanical, and electrical (HM&E) faults are sensors with stochastic components (e.g. temperature, pressure, etc.), this work focuses on providing the probability of observing a given sensor reading or fault based on prior measurements. Quantile regression forests (QRF) provide a means to identify outliers, produces measurement statistics, and handle the randomness associated with HM&E sensors. Given time, other AI/ML approaches may be compared to QRF. As many vessels have separate MCS for each subsystem, data engineering is needed to aggregate, synchronize, and decimate data as a preprocessing step. With no further preprocessing, preliminary work has found QRF to be useful to identity component failure. Future efforts plan to scale the training test set to aid in generating long-term trend and statistics and establishing a healthy baseline.

Application of AI/ML to Improve Unmanned Surface Vessel Machinery Control System Fault Handling

Autonomous systems, including self-driving cars and unmanned underwater/surface/ground/air/space vehicles, have caught the imagination of the press, the public, and personnel of the U.S. Department of Defense (DoD). Moreover, these systems increasingly harness the power of Artificial Intelligence and Machine Learning (AI/ML), more specifically Deep Learning (DL), for their operation. The stunning performance of deep learning algorithms in comparison to extant methods, including pattern matching, computational linguistics, statistical inferencing, and legacy machine-learning, has taken the world by storm. However, adoption of such systems in safety-critical applications has been the subject of intense debate and scrutiny, because of the propensity of these algorithms for erroneous operation, such as misdirection with seemingly innocuous signage, and their vulnerability to data poisoning and data sparsity attacks, leading to incorrect operation which can be both embarrassing and damaging. This has naturally led the DoD community to ask, “How do we harness this technology being unleashed upon the world, and yet keep our nation and warfighters safe?” Before we answer this question, however, it is important to note that trust is integral to DoD systems, including autonomous systems, and ensuring reliable system operations is paramount. Therefore, we need strategies to equip the DoD to design, build, deploy, and sustain autonomous systems that are trustworthy, secure, reliable, and dependable. In this paper, we investigate issues leading to poor performance and lack of robustness of autonomous systems based on machine learning, and discuss the state of the art for their mitigation.

Assuring Autonomous Cyber Physical Data Driven Systems (AA-CPDDS)

Autonomous Systems consist of different systems working together so they can sense, plan, and act to a dynamic environment. There is much confusion on what autonomous systems are. This session will explain what autonomous systems are and how machine teaching and simulation can collaborate to create autonomous systems. Machine Teaching is critical to autonomous systems. It is based on reinforcement learning. The differences between machine teaching and machine learning will also be clarified.

Understanding Autonomous Systems

Reinforcement learning (RL) is an effective method for solving problems that require autonomous agents to learn the best way to act in complex environments. RLlib is a powerful tool for applying reinforcement learning to problems where there are multiple agents or when agents must take on multiple roles. There exist many resources for learning about RLlib from a theoretical or academic perspective, but there is a lack of materials for learning how to use RLlib to solve your own practical problems. This tutorial helps to fill that gap. We show you how to apply reinforcement learning to your problem by taking you through a custom environment and demonstrating how to apply RLlib to that environment. We make the code available on GitHub.

Tutorial: RLlib for Deep Hierarchical Multiagent Reinforcement Learning

Although Convolutional Neural Networks (CNNs) achieve high accuracy on image recognition tasks, they lack robustness against realistic corruptions and fail catastrophically when deliberately attacked. Previous work suggests that CNNs with representations similar to primary visual cortex (V1) are more robust to adversarial attacks on images than current adversarial defense techniques. However, these approaches require training on large-scale neural recordings or handcrafting classical neuroscientific models. Motivated by long-standing evidence that neural activity in V1 and other sensory areas is sparse, we develop a class of hybrid CNNs, called LCANets, which feature a frontend with recurrent lateral competition. We demonstrate competitive accuracy of LCANets on action recognition datasets and show that LCANets are significantly more robust to image corruptions and adversarial examples than both standard CNNs and adversarially-trained CNNs. Our results demonstrate that recurrent lateral competition plays a large role in forming sparse and robust V1-like representations when incorporated into CNNs.

CNNs With a Neuromorphic Sparse Coding Frontend are More Robust to Adversarial Attacks

Today, essentially all computing is done with Boolean logic and binary numbers, requiring all manner of error correction codes to keep the most significant bits straight. But in the realm of machine learning, what if instead every bit was equally (in)significant? Consider two randomly generated strings thousands of zeros and ones long, about 50/50 split. Normally, about half the bits will be different between the two strings. If the number of differences is less than half though, the strings as a whole must be related. Even a large number of errant bits will not change the strings’ similarity. This creation and measurement of similarity between very large strings of bits (or numbers) is called hyperdimensional computing (HDC). 

Though not quantum mechanical, a kind of superposition of information across each of the thousands of bits enables machine learning capabilities even Deep Neural Networks (DNN) struggle with, such as learning from a single example or combining the learning from several different sensor modalities. One can even dynamically reprogram swarms of robots in a contested environment – the new program intelligible from a single binary string even with 30-50% noise.

What if we started training artificial neural networks (ANN) to learn and recall these strings: not memorizating or autoencoding but rather “resonating” like a tuning fork? This “resonance” training paradigm suddenly enables even more capabilities not possible with traditional DNNs today such as learning new categories “on the fly” without retraining the whole network from scratch. Such attractor state networks (ASN) lend themselves to modular design, allowing new categories to be added or shared amongst other ASNs. This combining of HDC with “resonance” ANNs is a potential game changer for machine learning, establishing mathematically rigorous engineering design rules instead of the unintelligible black boxes of DNNs today.

Hyperdimensional computing can be implemented in traditional computing hardware, but large size, weight, and power savings are possible by including memristors crossbars into the ML ensemble. The Pipeline for Novel Neuromorphic Hardware (NeuroPipe) Applied Research for Advancement of Priorities (ARAP) is working to bring not only novel neuromorphic computing hardware like memristors to the field but also develop the next generation of machine learning algorithms such as these.


NeuroPipe ARAP Hardware & Algorithms


Robots are increasingly deployed to solve complex real-world tasks that require  continuous processing of their observations only to be constrained by their on-board resources.  Energy-efficient control of mobile robots is one of the limiting factor in having robots deployed  to perform in-space assembly of structures. Spiking Neural Networks (SNNs) deployed on  neuromorphic processors provide an energy-efficient solution that could also facilitate building  intelligence similar to living creatures for robots. 
Neuromorphic architectures imitate the behavior and function of biological nervous  systems by implementing networks of artificial neurons and synapses that benefit from low  power consumption. These computing platforms emulate their biological counterparts by  distributing both computation and memory among a large number of computation units, each  passing information, via asynchronous spikes, to hundreds or thousands of other computational  units, like neurons would do through synapses. 
Over time, different digital neuromorphic architectures like Intel’s neuromorphic  technology, code named Loihi [1], have replicated many characteristics of SNNs in hardware.  Despite the advantages of neuromorphic hardware (e.g., high speed and energy efficiency)  there are few algorithms that utilize it for in situ robot control. Recent work by Michaelis et al. [2], proved the theoretical feasibility of using Intel’s neuromorphic technology for trajectory  generation on hardware, but fails to employ the strategy with robotic hardware (simulated or in  situ). Michaelis’ work presents a biologically plausible approach that also promised a control  relevant timescale.  
His work is based on Reservoir Computing, specifically, Liquid State Machines,  implemented as biologically plausible anisotropic networks. In Michaelis’ work for trajectory  generation, the liquid layer is implemented as an anisotropic network, based on a biologically  
inspired rule for forming spatially asymmetric non-plastic connections, similar to the ones that  have been observed to activate in human and mammal brains [3]. It is believed that correlated  spatial anisotropic connectivity is what makes learning faster and more efficient. Michaelis et 
Distribution A — Approved for public release, distribution is unlimited 
al. [2], make use of this anisotropic network idea from Spreizer et al. [3] and show the feasibility  of using an architecture that integrates it for trajectory generation.  
In this work we close the gap between theoretical formulation and practical applications  by implementing the missing components to enable Kinova Gen3, a 7 degree of freedom  robotic arm, to use this type of neural network on Intel Kapho Bay Loihi for trajectory  generation.  
Demonstrating neuromorphic capabilities for robotic control is of high interest at NRL  due to the appeal of high performance and energy efficiency that neuromorphic hardware  presents. The success of this strategy has a big impact on different projects at NRL that are investigating: neuromorphically controlled multi-limbed robotic systems for Naval ship  maintenance tasks, neuromorphically controlled multi-limbed space truss walking robot that  can autonomously assemble truss elements to construct large space structures in orbit, and the  capabilities of memristor-based neuromorphic processors. 

References: 

[1] Davies, M., Srinivasa, N., Lin, T.-H., Chinya, G., Cao, Y., Choday, S. H., et al. “Loihi: a  neuromorphic manycore processor with on-chip learning.” IEEE Micro (2018). 
[2] Michaelis, Carlo, Andrew B. Lehr, and Christian Tetzlaff. “Robust trajectory generation for  robotic control on the neuromorphic research chip Loihi.” Frontiers in neurorobotics 14 (2020). 
[3] Spreizer, Sebastian, Ad Aertsen, and Arvind Kumar. “From space to time: Spatial  inhomogeneities lead to the emergence of spatiotemporal sequences in spiking neuronal  networks.” PLoS computational biology 15.10 (2019).
Distribution A — Approved for public release, distribution is unlimited 
*

Premium content

Adaptive spiking control of a 7 DOF arm

Downloads

Next from NAML 2022

Performative Analysis of Sensor Limitations in Communications Denied Autonomous Swarms

Similar lecture

SpikePropamine: Meta-Learning Synaptic Plasticity for Spiking Neural Networks

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES