Hybrid Free-Space Optics/Radio Frequency (FSO/RF) Networks for Mobile Robot Teams

Consider a scenario where the infrastructure of a metropolitan area network (MAN) is incapacitated by a man-made or natural disaster (e.g., an earthquake). MAN connectivity could be automatically restored by a team of autonomous mobile agents if each were equipped with a communications medium capable of patching the broken links. However, in this case the link distance and throughput requirements might be of such magnitude to render radio frequency (RF) based approaches ineffective.

We predict that such throughput intensive scenarios will lead to the emergence of hybrid FSO/RF based mobile ad-hoc network (MANET) architectures. The benefits of such a model are immediately obvious. An FSO link in any network provides a high throughput channel over which time sensitive data can be transmitted. Many commercially available FSO devices already provide throughput rates in the Gbps range. Additionally, FSO provides constant throughput over a much longer range than RF-based channels. This is illustrated in the graph below, which compares throughput vs. range for free-space optics, ultra wide band (UWB), and 802.11x communication channels.

Throughput Graph

Agent Platforms

Our trials employed the use of two hybrid nodes (Balrog and Gimli) based upon the Pioneer P3-AT platform. Each agent is equipped with Laserwire 100 Mbps FSO transceiver mounted on a Pan/Tilt head, and an 802.11g RF interface. Both agents feature a Sony EVI-D70 18x PTZ camera system.

Robots/Agents

Hierarchical Link Acquisition

We propose the use of a three stage coarse-to-fine FSO link acquisition process (LAS). Coarse alignment is accomplished through the use of local sensor data, which provides each agent with a crude pose estimate in a common reference frame (we assume each agent has a priori knowledge of its link partner's destination position). Initial alignment is then refined using a vision-based robot identification and tracking system. This phase provides the information necessary in order to successfully align (using a PTZ mechanism) the respective optical transceivers, which is the final step in the process. Following are mosaic images acquired from a deployment of the LAS in Packard Laboratory. Each camera scanned a region +/- 15 degrees around its local axis of rotation. Click an image to enlarge.

Balrog Mosaic Thumbnail Gimli Mosaic Thumbnail

Packet Routing

The routing problem can be addressed by first observing that the FSO/RF MANET architecture has a natural physical network hierarchy with an FSO backbone at the highest layer and a collection of RF ad-hoc domains at the lowest. We propose the use of Hierarchical State Routing (Iwata, et al., 1999) as the primary means of routing within inter-domain RF address spaces. HSR lends itself nicely to the FSO/RF MANET architecture due to its support of differentiated service and node mobility. The routing approach in the backbone network will largely be a function of the number of nodes present at this level of the physical hierarchy. Assuming a modest number of hybrid nodes, traditional link-state routing can be employed. For larger systems, HSR can be applied to nodes comprising the optical backbone. Connectivity is ensured by having each hybrid node broadcast the subnetwork reachability information associated with its respective RF domain. Following is a diagram illustrating the distributed deployment of HSR in an FSO/RF physical hierarchical context.

FSO Physical Hierarchy

Experimentation

Our initial experiments to validate the FSO/RF model were designed to simulate a remote surveillance application. A pair of robots (Balrog and Gimli) were charged with establishing a FSO/RF MANET the length of a corridor in Packard Laboratory. In this scenario, the robots acquire a physical FSO link using the outlined hierarchical link acquisition system. Once a connection is established, the agents exchange video data across the FSO backbone. Balrog forwards data packets it receives over its RF interface to a Command Center where the user can view the data. In a similar manner, Gimli forwards the streaming video from Balrog over its RF interface to a designated notebook, which displays the image locally.

Network Diagram
Local Image Remote Image

Media

  1. Video Summery

Publications

People