Robotic Wireless Networks


There is growing agreement that wireless capacity (at the PHY and MAC layers) is reaching saturation. Many believe that the next “jump” in network capacity will emerge from new ways of organizing networks. While there exists substantial work on new network architectures, one assumption that most proposals seem to make is that infrastructure – WiFi APs, enterprise WLANs, cell towers – is static. This project considers the possibility of relaxing this assumption and explores the implications of physically moving wireless network infrastructure to improve/optimize desired performance metrics. For example, we envision WiFi access points on wheels that move within a small region to exploit the multipath nature of wireless signals; in the future, we envision drones flying into high demand areas, hovering at strategic locations, and serving as cellular proxies to ground clients. This project is a foray into the landscape of such "robotic wireless networks".



smart home




       People:





     Experimentation Platform:

platform

  Figure above shows an iMob AP assembled using a Roomba iRobot 2.1, a webcam, and a laptop equipped with  Intel 5300 802.11n cards. The laptop is mounted on the iRobot and connected to it over the serial interface; it is also connected to a Microsoft live cam (attached in front of the iRobot) to guide its motion. The laptop acts as the controller for the whole system, sending motion commands to the robot (via the OSI interface), while also controlling the network interface for transmission/reception. 8 laptop clients were uniformly scattered at various locations and programmed to communicate back to the iMob AP. The robot’s mobility is confined within a 2x2 feet square region, demarcated by colored duct tapes pasted on the floor. If the robot drifts out of the square box, the camera detects the color of the duct tapes and triggers a change in heading direction. The AP performs “raster scans” within the square box at a speed of 10 cm/sec – during the scan, the AP continuously sends around 200 packets/second, equivalent to 60 packets per 3cms. Transmissions are performed on regular OFDM with 3x3 MIMO at both 2.4GHz and 5GHz bands. Clients record the per-packet channel state information (CSI) for offline analysis.





         Main Results:
oracle

real gain
     




Publications:



 
          Future Directions:
Using drones as a cellular proxy. The key research challenge pertains to computing the
                                          location at which the drone should hover so that it maximizes, for example, the sume of
                                          SNR to all the clients that must connect to the drone. Searching for this location through
                                          through an efficient algorithm is non-trivial, however, ray tracing simulations may be
                                          used to guide the motion path of the drone.  The problem bears similarity to active
                                          learning.

         drone 





          UIUC/USC Collaboration:
  • Periodic brainstorming sessions and discussions with PI Nelakuditi and his team from USC. The discussions are mainly focussed on characterizing the practical aspects of Robotic Networks, namely the ramifications of imprecise motion of the robot, the limits of velocity, the difficulty in instantaneous braking, etc. These discussions have led to ironing out various pragmatic parts of the paper -- the findings and experiences from these exercises are being prepared for a submission to the IEEE Transactions of Mobile Computing. PI Roy Choudhury and Nelakuditi also submitted a CRI proposal to NSF.
  • Rufeng Meng, one of PI Nelakuditi's PhD students, visited UIUC from Summer 2015 to Fall 2015. He was collaborating on various aspects of mobility algorithms.
  • Ongoing collaboration with the USC team is focussing on drones and how they could serve as proxies to cellular towers, i.e., a drone flies into a region of high network congestion, positions itself strategically, and offers WiFi connectivity with cellular backhaul to the actual cell tower. Algorithmic questions pertain to the "search algorithm" so that the drone could find the best "hovering location".



    
          Educational and Outreach Activities:
 
  • Included in course material for "ECE/CS 498: Mobile Computing and Applications"
  • Invited seminar at the "Robotics Applications Workshop" at UC Berkeley
  • Planning on showcasing iMob at UIUC's Engineering Open House, 2017.
  • Short talk at NSF Workshop on Wireless Testbeds and Platforms




          Funding:
                               nsf funding         huawei