Real time GS scheduling based on multiagent

Transcription

1 Real time GS scheduling based on multiagent technology Dr. Anton B. Ivanov, EPFL Space Center, Switzerland Prof. Peter Skobelev, Samara State AeroSpace University / Smart Solutions Ltd. Vitaly Travin, Smart Solutions, Ltd, Samara Alexey Zhilyaev Smart Solutions, Ltd, Samara

2 Outline Problem description Multi-agent approach Software implementation Outlook for the future 2

3 EPFL Space Center (espace) Created early 2014, the EPFL Space Engineering Center is a spin-off of the Swiss Space Center and has the responsibility of all space projects and educational activities at EPFL. - Currently 7 people - Swiss Space Center has staff of 23 persons - Core engineering and management team since 2006 Our purpose is to foster space research and development projects, and design and implement innovative applications of space systems. We aim at nurturing a creative environment for students and engage wide collaborations among Swiss and International universities. The Space Engineering Center activities are funded by research programs (ESA, EU, CTI and others) as well as internal EPFL mandates. 3

4 Smart Solutions Multi-Agent Technology for Resource Management Brief info about Smart Solu1ons So4ware engineering company Smart Solu1ons established in 2010 (Samara, Russia) Part of Knowledge genesis Group of companies (started in 1990) Develops mul1- agent technologies for emergent intelligence systems designing based on fundamental principles of self- organiza1on and evolu1on which allow system con1nuously adapt to real 1me events and ever- changing opera1on condi1ons. Main fields: Aerospace Manufacturing Transport logis1c Mobile services Railways Supply chains More than 120 employees Annual turnover (mln. rub.): , , , More than 40 scien1fic publica1ons in 2013

5 Category Weight range, kg Example Nano 1-10 kg Cubesats Micro Brite-PL, Trailblazer Small SSTL-100, SSTL 150 Cubesat launch statistics Failed Failed (launch) Successful Unclear

6 Category Weight range, kg Example Nano 1-10 kg Cubesats Micro Brite-PL, Trailblazer Small SSTL-100, SSTL 150 Cubesat launch statistics Failed Failed (launch) Successful Unclear

7 Category Weight range, kg Example Nano 1-10 kg Cubesats Micro Brite-PL, Trailblazer Small SSTL-100, SSTL 150 Cubesat launch statistics Failed Failed (launch) Successful Unclear

8 Nanosats: Example constellations Flock 28 by planet labs - Goal: Distributed remote sensing, daily imaging. - Number of satellites: 28 3U (DOVE) satellites on the first NanoRacks launch - Ground station strategy: Proprietary network - Telecom band: VHF/UHF/S-band QB50 project ( to be launched in 2015 / 2016) - Goal: Atmospheric studies - Number of satellites: 50 2U and 3U satellites - Ground station strategy : GENSO baselined, but this is TBC - Telecom band: VHF/UHF Satellite altitude, km 430$ 380$ 330$ 280$ 230$ One of the Flock 28 satellites descent into atmosphere (object 39512) 180$ 10(Feb(2014$ 25(Feb(2014$ 12(Mar(2014$ 27(Mar(2014$ 11(Apr(2014$ 26(Apr(2014$ 11(May(201 Calendar date 6

9 CubETH: GNSS experiments in 1U Key science and technological objectives - Precise Orbit Determination (GPS, GLONASS and Galileo demo) - Experiments: radio occultations, reflectometry, air density - Satellite structure manufacturing using ALM (3D printing), easy assembly - Microelectronics demonstration: ultra low-power AOCS - Distributed ground stations for command and download Volume Cubesat Overview 1U (10x10x10cm) deg/s Angular norm speed > [days] BDOT On Gyro 2009 Gyro post 2011 Rate from audio SwissCube heritage: almost 5 years of operations Mass Power Data rate Payload Operations Orbit Launch 1.3 kg (per standard, TBC) 1.7 W (SwissCube) 9600 bps 11 ublox GNSS receivers 6 antennas Distributed Ground Stations 500km, SSO, 2AM/2PM 2016 (TBC) Payload Project Partners ETHZ, HSLuzern, HSRapperswil Systems Engineerin Swiss Space Center AIT Industrial partners Swiss Space Center ublox, RUAG, Saphyrion, CSEM Academic partners LEMA, RISD, 7

10 GS planning: Technical requirements. Network of automated ground stations shall consist of arbitrary stations. Network of stations shall receive arbitrary number uplink / downlink requests One request shall consist of - Name of the satellite - estimated size of requested data - latest time to download the data The network shall prepare and optimize the schedule to maximize efficiency (shortest time) or maximize load (minimize number of receiving stations). The system shall allow the following functions to - requests: new, edit, delete, schedule - stations: new, edit, delete, schedule The system shall allow the following constraints - individual satellite / station visibility - technical capabilities of individual stations 8

11 GS planning: Technical requirements. Network of automated ground stations shall consist of arbitrary stations. Network of stations shall receive arbitrary number uplink / downlink requests One request shall consist of - Name of the satellite - estimated size of requested data - latest time to download the data The network shall prepare and optimize the schedule to maximize efficiency (shortest time) or maximize load (minimize number of receiving stations). The system shall allow the following functions to - requests: new, edit, delete, schedule - stations: new, edit, delete, schedule The system shall allow the following constraints - individual satellite / station visibility - technical capabilities of individual stations 8

12 Approach: Multiagent technologies Definitions - agent is a autonomous object, which owns its own goals and capable to plan, decide and communicate depending on received interrupts / messages - multi agent system is a system, which contains many agents, placed in a common physical or virtual environment with a capability to satisfy a global optimization solution. Centralized System - Hierarchical subsystems - Consecutive operations - Top-down commanding - General stability - Rigidity Multiagent system - Large networks of small agents - Parallel operations - Interagent communication - Distributed solutions - Flexibility 9

13 Method We start with an initial set of conjugate agents sessions and stations, each of which has certain features or needs of other resources; We describe individual goals and criteria for decision making by all agents, as well as their preferences and constraints; We defined rules and protocols for interactions between agents that can detect conflicts and to find a compromise between the elements in establishing connections; We model coupled interactions inside our multi-agent system With this system, we construct the initial network needs and opportunities, determining the appropriate allocation of resources; - If the status of a resource or a demand changes with the arrival of new events, the network reoptimizes itself to resolve conflicts, but only in the part that is directly related to the changes; The system operation is completed, if network agents can not change its state, or it runs out of time allocated to the task scheduling. 10

14 Multiagent communication: algorithm Task agent Агент задачи Station agent Агент станции sat name, allowed interval CFP (имя спутника, допустимый интервал) i n refuse [Нет доступных для размещения интервалов] no available intervals n m propose (список доступных интервалов) j = n - i propose available interval reject-proposal j-1 Выбор варианта размещения из списка accept-proposal (интервал размещения) 1 failure [Выбранный интервал недоступен] Interval is not available inform [Задача запланирована] Task is planned

15 Modeling environment

16 Problem setup Select satellites of interest Assign tasks to the satellite Setup station constraints

17 Modeling Results Task: Each satellite to downlink 7.2 MB of data over 9.6 kbit/sec downlink Optimization for minimum time Optimization for minimum number of stations

18 Modeling Results Task: Each satellite to downlink 7.2 MB of data over 9.6 kbit/sec downlink Optimization for minimum time Optimization for minimum number of stations

19 Modeling Results Task: Each satellite to downlink 7.2 MB of data over 9.6 kbit/sec downlink Optimization for minimum time Optimization for minimum number of stations List of tasks per station

20 Modeling Results Task: Each satellite to downlink 7.2 MB of data over 9.6 kbit/sec downlink Optimization for minimum time Optimization for minimum number of stations List of tasks per station

21 Modeling Results Task: Each satellite to downlink 7.2 MB of data over 9.6 kbit/sec downlink Optimization for minimum time Optimization for minimum number of stations Visibility chart List of tasks per station

22 Conclusions and next steps Improvements in current version - better handling of satellite constellations (user interface) - realistic implementation of ground stations (constraints) - interface with automated station tracking software (e.g. HamRadio Deluxe) - quality tracking and automatic resending of missing data Use in projects - we approached the QB50 project to offer them downlink scheduler EPFL Space Center provides Mission Control Software (MCS) - this system will be used in the CubETH project for scheduling of data downlink at multiple stations Future 4 stations planned: Lucerne, Moscow, Samara, Krasnoyarsk. - this system has the capability to tracking quality of each individual agent and monetize downlink capabilities - it will be possible to offer to individual radio amateurs a refund for their activities to use their assets for downlink - expand this offer to commercial constellations to optimize downlink time and prioritize orders 15

23 Next step: Integration with CubETH Protocol Stack Space link OSI model: ground segment space segment (GSTS) MCS COM CDMS PL Appli- ca?on CCSDS/PUS TM/ /TC CCSDS/PUS TM/TC CCSDS/PUS TM/TC Internal func?on Internal func?on Transport Segment Network Packet Data Link Frame TCP IP Ethernet protocol TM/TC link protocol TM/TC link protocol IIC protocol IIC protocol IIC protocol IIC protocol TM/TC sync. + CH cod. TM/TC sync. + CH cod. Physical Bits Ethernet Radio frequency and modula=on IIC IIC