Michael D. Howard

Pheromone Robotics

We describe techniques for coordinating the actions of large numbers of small-scale robots to achieve useful large-scale results in surveillance, reconnaissance, hazard detection, and path finding. We exploit the biologically inspired notion of a “virtual pheromone,” implemented using simple transceivers mounted atop each robot. Unlike the chemical markers used by insect colonies for communication and coordination, our virtual pheromones are symbolic messages tied to the robots themselves rather than to fixed locations in the environment. This enables our robot collective to become a distributed computing mesh embedded within the environment, while simultaneously acting as a physical embodiment of the user interface. This leads to notions of world-embedded computation and world-embedded displays that provide different ways to think about robot colonies and the types of distributed computations that such colonies might perform.

Compound behaviors in pheromone robotics

Abstract We are pursuing techniques for coordinating the actions of large numbers of small-scale robots to achieve useful large-scale results in surveillance, reconnaissance, hazard detection, and path finding. Using the biologically inspired notion of “virtual pheromone” messaging, we describe how many coordinated activities can be accomplished without centralized control. By virtue of this simple messaging scheme, a robot swarm can become a distributed computing mesh embedded within the environment, while simultaneously acting as a physical embodiment of the user interface. We further describe a set of logical primitives for controlling the flow of virtual pheromone messages throughout the robot swarm. These enable the design of complex group behaviors mediated by messages exchanged between neighboring robots.

Distributed design review in virtual environments

In large distributed corporations, distributed design review offers the potential for cost savings, reduced time to market, and improved efficiency. It also has the potential to improve the design process by enabling wider expertise to be incorporated in design reviews. This paper describes the integration of several components to enable distributed virtual design review in mixed multi-party, heterogeneous multi-site 2D and immersive 3D environments. The system provides higher layers of support for collaboration including avatars, high fidelity audio, and shared artifact manipulation. The system functions across several interface environments ranging from CAVEs to Walls to desktop workstations. At the center of the software architecture is the Human Integrating Virtual Environment (HIVE) [6], a collaboration infrastructure and toolset to support research and development of multi-user, geographically distributed, 2D and 3D shared applications. The HIVE functions with VisualEyes software for visualizing 3D data in virtual environments. We also describe in detail the configuration and lessons learned in a two site, heterogeneous multi-user demonstration of the system between HRL Laboratories in Malibu, California and GM R&D in Warren, Michigan.

Fast pattern matching with time-delay neural networks

We present a novel paradigm for pattern matching. Our method provides a means to search a continuous data stream for exact matches with a priori stored data sequences. At heart, we use a neural network with input and output layers and variable connections in between. The input layer has one neuron for each possible character or number in the data stream, and the output layer has one neuron for each stored pattern. The novelty of the network is that the delays of the connections from input to output layer are optimized to match the temporal occurrence of an input character within a stored sequence. Thus, the polychronous activation of input neurons results in activating an output neuron that indicates detection of a stored pattern. For data streams that have a large alphabet, the connectivity in our network is very sparse and the number of computational steps small: in this case, our method outperforms by a factor 2 deterministic finite state machines, which have been the state of the art for pattern matching for more than 30 years.

Coalitions for distributed sensor fusion

We address the problem of efficient use of communication bandwidth in a network of distributed sensors. Each sensor node has enough computational power to fuse its own estimates with those of other nodes using an optimal filter, but message passing is expensive. The goal is to give each node an identical tactical picture of the area of interest without compromising target track accuracy. In the current leading approach, each sensing node sends an Associated Measurement Report (AMR: a sensor reading matched to a track ID) to every other node. This paper considers a method of decimation of reporting nodes by choosing a subset (coalition) of those nodes that have the best information on the target. Only coalition nodes share AMRs with each other; then the coalition sends a report to non-coalition nodes. Simulations and analytical studies are used to support the promise of the coalition approach.

Hippocampal anatomy supports the use of context in object recognition: a computational model

The human hippocampus receives distinct signals via the lateral entorhinal cortex, typically associated with object features, and the medial entorhinal cortex, associated with spatial or contextual information. The existence of these distinct types of information calls for some means by which they can be managed in an appropriate way, by integrating them or keeping them separate as required to improve recognition. We hypothesize that several anatomical features of the hippocampus, including differentiation in connectivity between the superior/inferior blades of DG and the distal/proximal regions of CA3 and CA1, work together to play this information managing role. We construct a set of neural network models with these features and compare their recognition performance when given noisy or partial versions of contexts and their associated objects. We found that the anterior and posterior regions of the hippocampus naturally require different ratios of object and context input for optimal performance, due to the greater number of objects versus contexts. Additionally, we found that having separate processing regions in DG significantly aided recognition in situations where object inputs were degraded. However, split processing in both DG and CA3 resulted in performance tradeoffs, though the actual hippocampus may have ways of mitigating such losses.

Amorphous Predictive Nets

This paper describes our approaches for coordinating the actions of extremely large numbers of distributed, loosely connected, embedded computing elements. In such networks, centralized control and information processing is impractical. If control and processing can be decentralized, the communications bottleneck is removed and the system becomes more robust. Since conventional computing paradigms provide limited insight into such decentralized control, we look to biology for inspiration.

The neural basis of decision-making during sensemaking: Implications for human-system interaction

We have created a high-fidelity model of 9 regions of the brain involved in making sense of complex and uncertain situations. Sense making is a proactive form of situation awareness requiring sifting through information of various types to form hypotheses about evolving situations. The MINDS model (Mirroring Intelligence in a Neural Description of Sensemaking) reveals the neural principles and cognitive tradeoffs that explain weaknesses in human reasoning and decision-making.

Adaptive recall in hippocampus

Complementary learning systems (CLS) theory describes how the hippocampal and cortical contributions to recognition memory are a direct result of their architectural and computational specializations. In this paper we model a further refinement of CLS that features separate handling of inputs from the dorsal and ventral posterior cortices, and present a possible mechanism for adaptive recall in hippocampus based on several research findings that have not previously been related to each other. This model suggests how we are able to recognize familiar objects in unfamiliar settings.

A computational model of perirhinal cortex: Gating and repair of input to the hippocampus

The medial temporal lobe-which includes the hippocampus, as well as perirhinal, parahippocampal, and entorhinal cortices-is required for declarative memory. We focus on the role of the perirhinal cortex (PRC) in relaying semantic representations from temporal cortex to the ventral hippocampus. It has been argued that the PRC is more than a simple relay. We review evidence that the PRC, in conjunction with the entorhinal cortex, serves to both gate information transfer to the hippocampus in response to an externally generated signal, and to improve the fidelity of this input prior to its mnemonic processing within the hippocampus. We present the first explicit model of externally mediated PRC gating based on several gating mechanisms previously modeled for generic cortical regions; we discuss the merits of our model with respect these existing theoretical gating mechanisms, and also outline three possible external control signals and their functional implications. We constructed a biologically plausible neural network model (InhibGate) based on the available literature, and compared it to a slightly adapted three layer network as a qualitative standard of comparison. In nearly every condition, the InhibGate network was more effective at information gating and pattern repair than the comparison three layer network. Our experiments support the proposal that inhibition within the rhinal cortices can block or admit information transfer to the hippocampus in response to an externally supplied excitatory signal. In addition, our experiments reveal a possible role for the rhinal cortices in repairing noisy or incomplete data: a role that has been previously ascribed to the hippocampus.