Olaf Witkowski | The University of Tokyo (original) (raw)

Papers by Olaf Witkowski

—Swarming is thought to critically improve the efficiency of group foraging, as it allows for err... more —Swarming is thought to critically improve the efficiency of group foraging, as it allows for error-correction of individual mistakes in collective dynamics. High levels of noise from the environment may require a critical mass of agents to make collective behavior emerge. It is therefore crucial to reach sufficient computing power to allow for these effects to emerge in simulations. We extend an abstract agent-based swarming model based on the evolution of neural network controllers, in order to explore further the emergence of swarming. Our model is grounded in the ecological situation in which agents can access some information from the environment about the resource location, but through a noisy channel. Swarming critically improves the efficiency of group foraging, by allowing agents to reach resource areas much more easily by correcting individual mistakes in group dynamics. As high levels of noise may make the emergence of collective behavior depend on a critical mass of agents, it is crucial to reach sufficient computing power to allow for the evolution of the whole set of dynamics in simulation. Since simulating neural controllers and information exchanges between agents is computationally intensive, in order to scale up simulations to model critical masses of individuals, the implementation requires careful optimization. We apply techniques from astrophysics known as treecodes to compute the signal propagation, and efficiently parallelize for multi-core architectures. Our results open up future research on signal-based emergent collective behavior as a valid collective strategy for uninformed search over a domain space.

Astrobiology, 2015

Contents 1. Introduction 1.1. A workshop and this document 1.2. Framing origins of life science 1... more Contents 1. Introduction 1.1. A workshop and this document 1.2. Framing origins of life science 1.2.1. What do we mean by the origins of life (OoL)? 1.2.2. Defining life 1.2.3. How should we characterize approaches to OoL science? 1.2.4. One path to life or many? 2. A Strategy for Origins of Life Research 2.1. Outcomes-key questions and investigations 2.1.1. Domain 1: Theory 2.1.2. Domain 2: Practice 2.1.3. Domain 3: Process 2.1.4. Domain 4: Future studies 2.2. EON Roadmap 2.3. Relationship to NASA Astrobiology Roadmap and Strategy documents and the European AstRoMap Appendix I Appendix II Supplementary Materials References.

From biological cells to bee swarms and bird flocks, nature shows countless examples of self-orga... more From biological cells to bee swarms and bird flocks, nature shows countless examples of self-organized groups displaying a collective mind. In such species, individuals interacting together end up producing an emergent behavior that increases their chances of survival and reproduction. This thesis shows an exploration of the evolution of communication through coordinated behaviors in populations of embodied agents. The goal is to reach a better under- standing of nature’s conditions for the evolution and strategies for the maintenance of collective behaviors. For that purpose, we present a framework making use of agent-based modeling to study the parallel evolution of coordination, cooperation and communication, for different types of interactions and levels of complexity. Through computer simulations, we test hypotheses on the conditions leading to synergistic behaviors and the evolution of honest communication. We first show signal-based swarming, in a population where the informa...

Species diversification is generally thought to emerge from space barriers, which isolate individ... more Species diversification is generally thought to emerge from space barriers, which isolate individuals from each other long enough so that they diverge significantly from each other, each adapting to their own ecological niche. However, the search for ecological opportunities alone does not account for all cases of speciation (Schluter 2000, Rundell & Price 2009). Signaling is observed between individuals of species having an advantage to commmunicate their identity via speciesspecific signals, leading to reproductive isolation (Nevo et al. 1987, Carlson & Arnegard 2011).

In Artificial Life, agent based modeling is a popular synthetic approach that often studies the e... more In Artificial Life, agent based modeling is a popular synthetic approach that often studies the evolutionary conditions responsible for adaptive group behavior. For example, emergent social phenomena such as communication and cooperation have been studied using agent models with a spatial distribution of agents and resources , . However, few studies have focused on the evolution of abstract concepts, such as a concept of time, that benefits individual and group behavior. In this study, agents attain a concept of time via learning to benefit from periodicity (cyclic resource growth) in the environment. Notable exceptions include the study of how memory extends an agents temporal horizon and increase its adaptability . Nehaniv (1999) discusses the concept of narrative intelligence in temporally grounded agents. For example, the impact that stories of the past have upon an agent group's social behavior. In related work, describe an information-theoretic model for individual and social learning in temporally grounded agents. The capacity to learn from environmental temporal patterns such as periodicity is beneficial to a broad spectrum of organisms, from Amoebae to human civilizations . This study investigates how an evolved sense of time can be used to adapt agent group behavior. The objective is to use a minimalist simulation model (with a spatial distribution of food and agents) to demonstrate that learning a concept of time facilitates efficient group foraging behavior. The concept of time is embedded into agent signals (indirectly indicating distances to food), and environmental behavior (seasonal variations define when food is scarce versus plentiful). Each agent is defined by a local clock (it's lifetime), and the environment by a global clock (oscillations of resource growth). The hypothesis is that resource growth cycles coupled with agent signaling about resource locations are sufficient conditions for agents to increase the efficiency of group foraging behavior. That is, agents adapt their behavior to exploit altruistic signals, learning when food is plentiful versus when it is not. presents an example of the environment (left) and the agent Artificial Neural Network (ANN) controller (right). Controllers are adapted with an Evolutionary Algorithm (EA) that evolves connection weights. Agent fitness equals the food amount consumed during a lifetime. Agents consume U energy units for standing still, and U + W energy units for moving. The EA selects for agent behaviors that stop and conserve energy when food is scarce, and behaviors that cause agents to move about foraging when food is plentiful. The environment is a two dimensional torus consisting of P evenly spaced food patches, governed by cyclic periods of food abundance (summer) and scarcity (winter). Each iteration, agents (speakers) emit a signal that conveys how many iterations in the past the speaker was on a food patch. From this, receivers (closest agents) learn that a food patch is Y grid spaces away in a given direction (agents receive signals from both directions).

Swarming behavior is common in biology, from cell colonies to insect swarms and bird flocks. Howe... more Swarming behavior is common in biology, from cell colonies to insect swarms and bird flocks. However, the conditions leading to the emergence of such behavior are still subject to research. Since Reynolds' boids, many artificial models have reproduced swarming behavior, focusing on details ranging from obstacle avoidance to the introduction of fixed leaders. This paper presents a model of evolved artificial agents, able to develop swarming using only their ability to listen to each other's signals. The model simulates a population of agents looking for a vital resource they cannot directly detect, in a 3D environment. Instead of a centralized algorithm, each agent is controlled by an artificial neural network, whose weights are encoded in a genotype and adapted by an original asynchronous genetic algorithm. The results demonstrate that agents progressively evolve the ability to use the information exchanged between each other via signaling to establish temporary leader-follower relations. These relations allow agents to form swarming patterns, emerging as a transient behavior that improves the agents' ability to forage for the resource. Once they have acquired the ability to swarm, the individuals are able to outperform the non-swarmers at finding the resource. The population hence reaches a neutral evolutionary space which leads to a genetic drift of the genotypes. This reductionist approach to signal-based swarming not only contributes to shed light on the minimal conditions for the evolution of a swarming behavior, but also more generally it exemplifies the effect communication can have on optimal search patterns in collective groups of individuals.

It brings out the animal in us " is often heard, when speaking of unaltruistic behavior. Frans de... more It brings out the animal in us " is often heard, when speaking of unaltruistic behavior. Frans de Waal has argued against a " veneer theory " of one of humanity's most valued traits: morality. It has been proposed that morality emerges as a result of a system of evolutionary processes, giving rise to social altruistic instincts. Traditional research has been arguing that fully-fledged cognitive systems were required to give each individual its autonomy. In this paper, we propose that a simple sense of morality can evolve from swarms of agents picking actions such that they are viable to the survival of the whole group. In order to illustrate the emergence of a moral sense within a community of individuals, we use an asynchronous evolutionary model, simulating populations of simulated agents performing a foraging task on a two-dimensional map. We discuss the morality of each emergent behavior within each population, then subsequently analyze several cases of interactions between different evolved foraging strategies, which we argue bring some insight on the concept of morality out of a group, or across species.This proposed approach brings a new perspective on the way morality can be studied in an artificial model, in terms of adaptive behavior, corroborating the argument in which morality can be defined not only in highly cognitive species, but across all levels of complexity in life.

Swarming behavior is common in biology, from cell colonies to insect swarms and bird flocks. Howe... more Swarming behavior is common in biology, from cell colonies to insect swarms and bird flocks. However, the conditions leading to the emergence of such behavior are still subject to research. Since Reynolds' boids, many artificial models have reproduced swarming behavior, focusing on details ranging from obstacle avoidance to the introduction of fixed leaders.
This paper presents a minimalistic agent-based model, in which individuals develop swarming using only their ability to listen to each other's signals.
The model simulates a population of agents looking for a vital resource they cannot directly detect, in a 3D environment. Each agent is controlled by an artificial neural network, whose weights are encoded in a genotype and evolved by an original asynchronous genetic algorithm.
The results demonstrate that agents progressively become able to use the information exchanged between each other via signaling to establish temporary leader-follower relations. These relations allow agents to form swarming patterns, emerging as a transient behavior that improves the agents' ability to forage for the resource. Once they have acquired the ability to swarm, the individuals are able to outperform the non-swarmers at finding the resource. The population hence reaches a neutral evolutionary space which leads to a genetic drift of the genotypes.
This reductionist approach to signal-based swarming not only contributes to shed light on the minimal conditions for the evolution of a swarming behavior, but also more generally it exemplifies the effect communication can have on optimal search patterns in collective groups of individuals.

The evolution of cooperation has long been studied in Game Theory and Evolutionary Biology. In th... more The evolution of cooperation has long been studied in Game Theory and Evolutionary Biology. In this study, we investigate the impact of movement control in a spatial version of the Prisoner's Dilemma in a three dimensional space. A population of agents is evolved via an asynchronous genetic algorithm, to optimize their strategy. Our results show that cooperators rapidly join into static clusters, creating favorable niches for fast replications. Surprisingly, even though remaining inside those clusters, cooperators keep moving faster than defectors. We analyze the system dynamics to explain the stability of this behavior.

The spontaneous generation of life has long been a central question investigated in the study of... more The spontaneous generation of life has long been a central question investigated in the
study of the origins of life [Szathmáry and Smith, 1995, Bedau et al., 2000]. The most
common constructive approach to this problem might be artificial chemistry, the computer-
inspired modeling of systems composed of chemical substances, either simulated with
interaction rules and with more or less coarse-grained structures or implemented in vitro.
Reaction–diffusion (RD) systems, first introduced by Alan Turing [Turing, 1952], are ...

Computational modeling is an important tool in the study of language evolution. It is not only us... more Computational modeling is an important tool in the study of language evolution. It is not only used to test hypotheses, but also as a source of data on difficult to observe evolutionary dynamics. This makes it particularly important to distinguish the emergent behaviors of evolutionary systems being studied, from the behaviors of specific models. In this paper we provide an in-depth analysis of one recent model of linguistic bio-cultural coevolution (Yamauchi & Hashimoto 2010) and show that several of its reported behaviors are artifacts produced by the model’s design and parameter settings. Specifically, we show that the model’s population size setting and agent “geography” place strong limits on both cultural and biological diversity in the model. These limits interact with the model’s learning mechanism and result in a number of semi-stable attractor states. We argue that it is the proper- ties of these attractors that account for the long run behavior of the model, directly conflicting with the analysis given in the original paper. Our results are confirmed by experiments altering the model’s population size parameter which result in a qualitative change in the observed model behavior.

In Evolutionary Biology and Game Theory, there is a long history of models aimed at predicting st... more In Evolutionary Biology and Game Theory, there is a long history of models aimed at predicting strategies adopted by agents during resource foraging. In Artificial Life, the agent-based modeling approach allowed to simulate the evolution of foraging behaviors in populations of artificial agents embodied in a simulated environment. In this paper, different sets of behaviors are evolved from a simple setting where agents seek for food patches distributed on a two-dimensional map. While agents are not explicitly playing a game of chicken, their strategies are found on a spectrum ranging from a frugal strategy (aka Dove) to a greedy strategy (aka Hawk). This phenomenon is due to the fact that moving is both a way for the agents to play or go to get away from an unfavorable area of the environment. It is also observed that by moving away, the agents preserve the ecology, preventing the resource from disappearing locally. Those strategies are shown to be stable if the environment is colonized by one given population. However, post-mortem tournaments among different groups of agents (separately evolved), systematically result in a specific group of agents dominating. The optimal strategy in the simulated tournaments is found to be one with fine-tuned timing for leaving. Further analysis shows how the strategy exploits resources without completely depleting them, producing Volterra-like population tendencies.

Since Reynolds boids, swarming behavior has often been reproduced in artificial models, but the c... more Since Reynolds boids, swarming behavior has often been reproduced in artificial models, but the conditions leading to its emergence are still subject to research, with candidates ranging from obstacle avoidance to virtual leaders. In this paper, we present a multi-agent model in which individuals develop swarming using only their ability to listen to each others signals. Our model uses an original asynchronous genetic algorithm to evolve a population of agents controlled by artificial neural networks, looking for an invisible resource in a 3D environment. The results demonstrate that agents use the information exchanged between them via signaling to form temporary leader-follower relations allowing them to flock together.

In nature, animals rely upon migratory behaviors in order to adapt to seasonal variations in thei... more In nature, animals rely upon migratory behaviors in order to adapt to seasonal variations in their environment. However, the transmission of migratory behaviors within populations (either during lifetimes or throughout successive generations) is not well understood (Bauer et al., 2011). In Artificial Life research, Agent Based Modeling (ABM) is a bottom-up approach to study evolutionary conditions under which adaptive group behavior emerges. ABM is characterized by synthetic methods (understanding via building), and is becoming increasingly popular in animal behavior research (Sumida et al., 1990). Combining an Artificial Neural Network (ANN) and Evolutionary Algorithm (EA) for adapting agent behavior (Yao, 1993) has received significant research attention (Phelps and Ryan, 2001), (Lee, 2003).

Astrobiology, 2015

Swarming behavior is common in biology, from cell colonies to insect swarms and bird flocks. Howe... more Swarming behavior is common in biology, from cell colonies to insect swarms and bird flocks. However, the conditions leading to the emergence of such behavior are still subject to research. Since Reynolds' boids, many artificial models have reproduced swarming behavior, focusing on details ranging from obstacle avoidance to the introduction of fixed leaders.
This paper presents a minimalistic agent-based model, in which individuals develop swarming using only their ability to listen to each other's signals.
The model simulates a population of agents looking for a vital resource they cannot directly detect, in a 3D environment. Each agent is controlled by an artificial neural network, whose weights are encoded in a genotype and evolved by an original asynchronous genetic algorithm.
The results demonstrate that agents progressively become able to use the information exchanged between each other via signaling to establish temporary leader-follower relations. These relations allow agents to form swarming patterns, emerging as a transient behavior that improves the agents' ability to forage for the resource. Once they have acquired the ability to swarm, the individuals are able to outperform the non-swarmers at finding the resource. The population hence reaches a neutral evolutionary space which leads to a genetic drift of the genotypes.
This reductionist approach to signal-based swarming not only contributes to shed light on the minimal conditions for the evolution of a swarming behavior, but also more generally it exemplifies the effect communication can have on optimal search patterns in collective groups of individuals.