Ryo Kanamori

Evaluation of anticipatory stigmergy strategies for traffic management

In this paper, we propose and evaluate an anticipatory stigmergy model for traffic management. Recently, dynamic coordination methods are becoming possible using the more short-term traffic information that can be provided by probe-vehicle information or smart phones. Some approaches have been trying to handle short-term traffic information in which a stigmergy-based approach is employed as an indirect communication method for cooperation among distributed agents and for managing traffic congestion. One drawback of these approaches is that handling near-future congestion remains problematic because stigmergies are basically past information. We propose anticipatory stigmergy for sharing information on near-future traffic and allocating drivers adequately. In this model, all vehicles submit their near-future intention as anticipatory stigmergy to search again their routes. Because it might be ineffective that all drivers choose the recommended route by anticipatory stigmergy, we examine strategies to allocate drivers the recommended link based on the residual distance or the lost time of traffic congestion. Our preliminary results demonstrate that anticipatory stigmergy with allocation strategy works well.

The comparison of stigmergy strategies for decentralized traffic congestion control: preliminary results

We investigate several stigmergies models for decentralized traffic congestion control. For realizing a smart city, one of the main problems that should be handled is traffic congestion. There have been a lot of works on managing traffic congestion with information technology. There is a relatively long history on observing traffic flow and then providing stochastic estimation on traffic congestion. Recently, more dynamic coordination methods are becoming possible by using more short term traffic information. Short term traffic information can be provided by car navigation systems with GPS (Global Positioning System)s and probe-vehicle information. There are several approaches to handle short term traffic information, in which stigmergy-based approach is a popular. Stigmergy is employed for indirect communication for cooperation among distributed agents. We can imagine several types of stigmergies : long term memory, short term memory, and anticipatory memory. However, there have been no discussion what kind of stigmergies can work well for managing traffic congestion. We conducted several simulations to compare the different kind of stigmergies. Our preliminary results demonstrate that if the traffic network is static, the combination of long term and short term stigmergies overcome the other stigmergies.

Auction-Based Parking Reservation System with Electricity Trading

Parking reservations can reduce the amount of time lost by drivers searching for parking spaces near their destination. For the development and planning of smart cities with electric vehicles, it is important to control parking spaces as places for charging and power management with Vehicle-to-Grid systems. This study proposes an auction-based parking reservation system that includes electricity trading, and uses simulation combined with a driver parking duration model to evaluate it. The driver parking duration model is constructed using actual parking data that can estimate parking times after parking fees have changed.

Agent-based electrical power management model for houses equipped with storage battery and photovoltaic units

Smart grid systems have been actively discussed to realize a sustainable and a low-carbon society that efficiently consumes electric power and to introduce photovoltaic power generation, i.e., renewable energy or electric vehicles. In this study, we focus on the smart meter system based on an agent model to manage electric power use in residential homes with which solar panels and a storage battery are equipped. We evaluated the system for one month during weather changing patterns.

Evaluation of Automated Negotiation System for Changing Route Assignment to Acquire Efficient Traffic Flow

We propose a negotiation system for changing route assignment to acquire efficient traffic flow. In this system, each probe-vehicle that is assigned a detour based on anticipatory stigmergy requests another probe-vehicle that is assigned a route based on the conventional approach to change route assignment. Anticipatory stigmergy collects near-future traffic positions of probe-vehicles. Recently, advances in Intelligent Transport Systems suggest a future in which vehicles will handle their own positions. Our foresight-route-providing method combines a previous method that utilized the past travel times of vehicles for decentralized traffic congestion management. We assume that all vehicles collect dynamic traffic conditions in decentralized processing environments as real implementation environments during the simulations and negotiate among vehicles based on a rational probe vehicle. In this paper, we demonstrate that an automated negotiation system has the potential for significant efficiency and stability for traffic flow in decentralized processing environments that can collect dynamic traffic conditions from the previous few minutes and estimate the vehicle positions after a few minutes in only limited area. We confirm that an automated negotiation system for changing route assignment contributes to the efficiency of traffic flow and time loss in congestion of each probe vehicle.

Stability Evaluation of Route Assignment Strategy by a Foresight-Route under a Decentralized Processing Environment

We propose a foresight-route-providing method based on anticipatory stigmergy for collecting near-future traffic positions. Even though traffic flow information is better than no traffic information, drivers still face dangerous congested traffic flow. Several approaches have addressed short-term traffic information in which a stigmergy-based approach is employed as an indirect communication method for cooperation among distributed agents and for managing traffic congestion. Recently, advances in Intelligent Transport Systems suggest a future in which vehicles handle their own positions. Our foresight-route-providing method combines a previous method that utilized the past travel times of vehicles for decentralized traffic congestion management. We assume that all vehicles collect dynamic traffic conditions in decentralized processing environments as real implementation environments during the simulations. In our simulations, we demonstrate that our assignment route strategies have the potential for significant efficiency and stability for traffic flow in decentralized processing environments that can collect dynamic traffic conditions from the previous few minutes and estimate the vehicle positions after a few minutes in only limited links. In this paper, we evaluate the differences between centralized and decentralized processing environments and investigate how the ratio of probe vehicles that are equipped with such devices as navigation systems is related to efficient and stable traffic flow. We confirmed that a route assignment strategy contributes to the stable efficiency of the total travel time. With more probe vehicles, assignment route strategies becomes efficient in our approach.

Evaluation of special lanes as incentive policies for promoting electric vehicles

In this study, we evaluate the effects of electric vehicle (EV) lanes in which only EV drivers enjoy the benefits of a reduction of travel time like the HOV lanes, and EV/Toll lanes whose charging system is similar to HOT lanes. The multi-class combined equilibrium model including a nested logit model to describe the travelers behavior of each EV owners and non-owners is developed. The differences in service levels between EV drivers and non-EV drivers are not only a travel time under the special network for EV but also travel costs. If EV diffusion rate in the Nagoya metropolitan area is 10% and EV/Toll lanes are introduced, the incentive for owning EV is ensured, and the social surplus is improved than before the introduction of special lanes.

One Cycle of Smart Access Vehicle Service Development

Under JST RISTEX S3FIRE program, we are trying to implement Smart Access Vehicle (SAV) Service in Hakodate. The project adopts the method of service science loop – the repeated cycle of observation, design and implementation. In this paper we report the completion of its first cycle, and discuss how the cycle improved our initial design. We first conducted person trip research in Hakodate. We chose 20 candidates of various age and occupation, and recorded their everyday movements for 4 months. We then analyzed the result and made a person trip model. The model was then fed into our multi-agent simulator for Hakodate public transportation system. We conducted a small field test with five vehicles for 1 week. The most significant achievement is that we confirmed that our design of SAV system works. We succeeded in automatically dispatching five vehicles for 11 h without any significant trouble or human supervision.

Application of incentive based scoring rule deciding pricing for smart houses

This paper presents a smart pricing scheme for smart house facilitating a scoring rule operated reward based pricing mechanism. In this scheme, provider (EP) monitors the network load and propose a day-ahead pricing to the consumers. The consumers respond that pricing by providing a probabilistic device schedule prediction for the smart devices to the lower period. The EP incentivizes the consumers by offering rewards (or discount) over the price depending on the accuracy margin of truthfulness of shifting-period forecast. Such reward function is formed based on a Continuous Ranked Probability Score (CRPS). CRPS has the ability to be a strictly proper scoring rule as well as to assess the closeness of estimation. Finally an optimization problem is formed which reduces EPs costs of providing rewards and satisfying consumers demand. The simulation results will show the effectiveness of the proposed method.

Evaluation of Community-Based Electric Power Market with Multi-agent Simulation

Electric power trading has been adopted in many countries. In this paper, we focus on community-based electricity markets to reduce a demand of electric power and ensure an balancing supply of energy for renewable energy. The community can be defined as a domestic group of individual homes. We assume a community has a larger battery that can be shared by individual homes. In the real world, community-based battery sharing has been focused on as a useful electricity buffer in crisis situations and also for stabilizing usage of electricity in a domestic area. In this paper, we build several domestic electric power management agent models by using Japanese power consumption data and meteorological data. We evaluate community-based electricity markets by using agent-based simulation, in which the validity of the large-sized storage battery in a community is confirmed regarding the load mitigation to a power network.