Abdul-Halim Jallad, Tanya Vladimirova Page 1 MAPLD 2005/1005 Operating Systems for Wireless Sensor Networks in Space Abdul-Halim Jallad and Tanya Vladimirova

Abdul-Halim Jallad, Tanya Vladimirova Page 2 MAPLD 2005/1005 Outline of Presentation Applications of wireless sensor networks in space Formation flying missions overview Requirements analysis of operating systems for formation flying missions Testbed development Conclusions

Abdul-Halim Jallad, Tanya Vladimirova Page 3 MAPLD 2005/1005 Wireless Sensor Networks: Convergence of Technologies Sensors: Miniaturization and micromachining makes tiny and low- cost sensors available commercially Embedded computing: Small and low-cost processors that are networked together facilitate collaboration through information and resource sharing Wireless communications: optical and RF communications enable networking between nodes Wireless sensor networks

Abdul-Halim Jallad, Tanya Vladimirova Page 4 MAPLD 2005/1005 Wireless Sensor Networks in Space 1) Manned Spacecraft missions: e.g. crew health monitoring Temperature Sensors 3) Spacecraft Diagnostics and monitoring 4) Inter-planetary Exploration Figure from 2) Spaced-based formation flying wireless sensor networks

Abdul-Halim Jallad, Tanya Vladimirova Page 5 MAPLD 2005/1005 Multi-Satellite Missions: Terminology A Virtual Satellite is a spatially distributed network of individual satellites collaborating as a single functional unit, and exhibiting a common system-wide capability to accomplish a shared objective. A Distributed Space System (DSS) is a system that consists of two or more satellites that are distributed in space and form a cooperative infrastructure for science measurement data acquisition, processing analysis and distribution. A Sensor Web is a system of intra- communicating spatially distributed sensor crafts that may be deployed to monitor environments. Sensor webs may involve many non-space elements and are therefore not completely covered by DSS. A Constellation is a group of satellites that have coordinated coverage, operating together under shared control, synchronised so that they overlap well in coverage and reinforce rather than interfere with other satellites' coverage. A Cluster is a functional grouping of spacecraft, formations, or virtual satellites. A Formation is a multiple-spacecraft system with desired position and/or orientation relative to each other or to a common target. Formation flying is the term used for the tracking and maintenance of a desired relative separation, orientation or position between or among spacecraft.

Abdul-Halim Jallad, Tanya Vladimirova Page 6 MAPLD 2005/1005 Formation-Flying Missions: Types Signal Separation: Measurements from the same source are collected by spatially distributed sensors on- board different nodes in the formation e.g. large synthetic apertures. Signal Combination: Distinct sensors on separate nodes collect data from different sources and merge this data on-board of the formation to extract global information of a particular phenomenon e.g. Earth observation-1 mission. Signal Coverage: A Sensor Web with identical sensors on the nodes with the purpose of covering wide areas of surface (e.g. multi-point sensing).

Abdul-Halim Jallad, Tanya Vladimirova Page 7 MAPLD 2005/1005 Formation-Flying Missions: The Information System Formation- Flying Missions: Information System Sensors and Actuators: These may be divided into three classes – spacecraft specific, formation-flying specific and payload specific On-Board Computing: Hardware is to be power and memory efficient while being fault-tolerant. Software includes: – mission software – middleware – an operating system to support distributed services. Inter Satellite Communications: Intersatellite links are different from terrestrial WSN wireless links in two main aspects: large distances involved and predictability

Abdul-Halim Jallad, Tanya Vladimirova Page 8 MAPLD 2005/1005 Model Application To investigate the advantages and disadvantages of distributed computing on-board of formation-flying (FF) missions To study possible implementations of distributed computing on-board FF missions To propose an optimal operating system architecture for such missions For the purpose of narrowing down the scope of this investigation we focus on a particular type of FF missions – virtual satellites Application: Sensor web: Imaging Signal Separation: Synthetic apertures The satellite nodes: Mass <= 1 Kg Area <= 1 cm3 Power <= 2 Watts Orbit = Low Earth Orbit (LEO) ~ 600Km Mission Model Aims of Research The Network Separation distances = in the order of kilometers Use of directional antennas.

Abdul-Halim Jallad, Tanya Vladimirova Page 9 MAPLD 2005/1005 Formation-Flying Mission: Information System Architecture System Threads Address space Files Hardware Drivers Physical Data Link Network Transport Sensor Driver HardwareSensor Middleware management AlgorithmsModules Services Virtual Machine App1App2App3 P o w er M a n a g e m e nt Application Hardware Middleware Operating System

Abdul-Halim Jallad, Tanya Vladimirova Page 10 MAPLD 2005/1005 OS Design for Formation-Flying Missions Process description and control:  Fault-tolerance: e.g. process replication  Memory considerations Concurrency:  FF missions are distributed systems and involve concurrency Memory management:  Use of bulk memory  Program memory wash Input/output management File management:  Fault-tolerance Networking:  Space protocol for ISL and ground space links Security Scheduling:  Real-Time scheduling  Low-power scheduling Process Description and Control Concurrency Networking File Management Security Input/Output Management Scheduling Memory Management Main Functions:

Abdul-Halim Jallad, Tanya Vladimirova Page 11 MAPLD 2005/1005 OS Design Factors for Formation-Flying Missions OBDH  The architecture of the on-board data handling system (e.g. distributed, centralized, multi- processor etc.) affect the operating system design ISL  The OS needs to consider the bandwidth, power consumption and unreliability of the inter- satellite links while making distributed decisions Formation Flying (FF)  The effect of the relative dynamics brought by FF on the OS design needs to be investigated On-board Software  The nature of the applications running on-board and its distribution among the FF nodes may have a direct impact on the OS design Constraints  The limited size and therefore available energy for computation and communication is an important factor that the OS design has to consider Factors Operating System

Abdul-Halim Jallad, Tanya Vladimirova Page 12 MAPLD 2005/1005 On-Board Data Handling for Pico-Satellites OBDH Ultra-low Power Advanced Packaging Reconfigurable hardware SOC* ASICs FPGAs SiGe on SOI * = system-on-a-chip: may involve various technologies including mixed-signals (analog/digital) on a single substrate Multi-processor Systems Time-Scale = ???

Abdul-Halim Jallad, Tanya Vladimirova Page 13 MAPLD 2005/1005 Types of Operating Systems Operating System DescriptionProsConsExample/ Mission Monolithic Almost any procedure can call any other procedure. EfficientLack modularity OS: Linux Mission: None Microkernel (client/server) A few essential functions are embedded in the kernel. Other services run as processes in user mode. Flexible Well suited for distributed systems Less efficient than monolithic OS: QNX, VxWorks Missions: TiungSAT-1, PROBA Virtual Machines Exact copy of bare hardware.PortableLow- performance OS: Embedded Java Virtual machine Mission: None Component- Based The Operating system consists of a set of independent components representing system resources Portable Efficient Well suited for distributed systems OS: TinyOS Mission: None

Abdul-Halim Jallad, Tanya Vladimirova Page 14 MAPLD 2005/1005 The TinyOS: Component-Based OS Operating system specifically designed for wireless sensor networks Applications consist of scheduler and a graph of components “Higher-level” components issue commands to and respond to events from “Lower-level” components Components contain: Set of command handlers, Set of event handlers, A fixed size storage frame, Collection of simple threads which can be scheduled. TinyOSTinyOS Component Components can be implemented in hardware or software. Events propagate upward in the hierarchy Commands propagate downward in the hierarchy. TinyOS Application Frame Tasks Commands received Events received Events initiated Commands made

Abdul-Halim Jallad, Tanya Vladimirova Page 15 MAPLD 2005/1005 Operating System Design for Swarms of Pico-Satellites Fault tolerance Small foot-print Low-power consumption Support for reconfigurable computing. Distributed system support Scalability Support for inter-satellite link communications Thread-based model Event-based model Conclusion: The component-based structural model provides flexibility, reusability and is suitable for distributed systems design while the event- based behavioural model provides speed, low power and memory efficiency. Design RequirementsComponent- Based Model Execution-Model Component library -Tasks perform computations -Tasks are implemented as finite state machines - States of tasks are transitioned through events -The system uses a main thread, which hands off tasks to individual task- handling threads -High context switch overhead

Abdul-Halim Jallad, Tanya Vladimirova Page 16 MAPLD 2005/1005 Distributed Computing for Formation-Flying Missions: Testbed GR-PCI-XC2V-FT LEON-3 Multiprocessor OBC XSV800 LEON-3 Multiprocessor OBC XSV800 LEON-3 Multiprocessor OBC Ethernet Windows XP PC STK Matlab Simulink Satellite Tool Kit TCP/IP server STK Advanced AO STK/ Connect Linux development platform DSU Monitor DDD GCC Compiler Programming Environment RS232 Visualization

Abdul-Halim Jallad, Tanya Vladimirova Page 17 MAPLD 2005/1005 System Emulation GR-PCI-XC2V-FT XC2V3000 Virtex-II FPGA Ethernet PHY interface LEON-FT core Support On-board memory SRAM SDRAM Flash PROM Figure from the “ LEON-PCI-XC2V Development board user manual ” XSV800 XCV800 Virtex FPGA Ethernet PHY interface On-board memory SRAM Flash Prom Figure from the website Mica2 motes 916MHz Multi- channel Radio Transceiver ATMEL128L 8-bit low-power processor Compatible with TinyOS (specifically designed for sensor networks). Node Emulation Hardware Distributed System Emulation Hardware Figures from mica2 datasheet

Abdul-Halim Jallad, Tanya Vladimirova Page 18 MAPLD 2005/1005 Pico-Satellite Computing Platform The chosen processor is the LEON-3 soft IP core 32-bit SPARC V 8 architecture Could be used in a multi- processor system Soft core (suitable for developing system-on-chip prototypes) Power-down mode is supported Embedded Hardware Debug Support Unit (DSU). LEON-3 in a multi-prosessor configuration Figure from

Abdul-Halim Jallad, Tanya Vladimirova Page 19 MAPLD 2005/1005 Conclusions Wireless sensor networks are a promising technology for space applications including orbital formation-flying (FF) missions and inter-planetary exploration. This research focuses on implementation of distributed computing on-board FF missions employing the wireless sensor networks concept. The various factors that affect the operating system (OS) design of FF missions may be divided into two categories: Traditional OS requirements: e.g. code efficiency and real-time performance. Specific requirements for FF missions: e.g. fault-tolerant distributed computing, orbit dynamics etc. A novel OS for multi-satellite FF missions should have the following features: An event-based execution model allowing to achieve low-power consumption and to fulfil the concurrency requirement with minimal amount of code. A component-based structural model allowing to achieve the modularity requirement and enabling the hardware/software boundary crossing, which provides support for reconfigurable and distributed computing. The TinyOS is selected as the baseline OS to be studied and adapted for use in distributed FF satellite missions.