[net2-wg] Implementation of the ZigBee network layer - open-ZB implementation status - André Cunha

[André Cunha]

Hello all,
I want to thank all the members of the net2-wg for this participation 
opportunity.
The following status report is a summary of our work on the IEEE 
802.15.4/ZigBee.
Best Regards

André Cunha

*
*

*
*

*Open-ZB implementation status *

*ZigBee Network Layer*

-Implementation of the SAPs interfaces

-Data transfer between NWK and MAC

-Association mechanisms

-ZigBee Addressing schemes

-Tree-routing

We also proposed a Time Division Beacon Scheduling approach that tackles 
the beacon synchronization problem in cluster tree networks. (See 
http://www.open-zb.net/publications/hurray-tr-070102.pdf for a Technical 
Report on some aspects of the implementation of this mechanism and some 
details of the ZB NWK implementation). Two papers on this mechanism are 
going to be presented in two international conferences.

*IEEE 802.15.4 MAC and PHY*

Functionalities already implemented

- SAP interfaces

- CSMA/CA algorithm – slotted version;

- GTS Mechanism;

- Indirect transmission mechanism;

- Direct / Indirect / GTS Data Transmission;

- Beacon Management;

- Frame construction – Short Addressing Fields only and extended 
addressing fields in the association request;

- Association/Disassociation Mechanism;

- MAC PIB Management;

- Frame Reception Conditions;

- ED and PASSIVE channel scan;

Functionalities not yet implemented

- Unslotted version CSMA/CA – Implemented but not fully tested;

- Extended Address Fields of the Frames;

- IntraPAN Address Fields of the Frames;

- Active and Orphan channel Scan;

- Orphan Devices;

- Frame Reception Conditions (Verify Conditions);

See http://www.open-zb.net/publications/tr-061106.pdf for implementation 
details of the IEEE 802.15.4

*Summary of the implementation problems and difficulties*

-In order to maintain the IEEE 802.15.4 (beacon intervals, superframe 
durations, time slots and backoffs) we had to implement an asynchronous 
timer based on the Micaz hardware clock. The different timer 
granularities in heterogeneous mote networks can be an implementation 
problem.

-We experienced that long processing leads to a de-synchronization of 
the devices (MicaZ)

-Memory constraints: either for the buffer management and for the 
routing tables (neighbours, network discovery and routing). We feel that 
the MicaZ available RAM memory is not enough to store and maintain all 
this data.

-Physical layer turnaround time (TX to RX and vice-versa) of the CC2420 
plus the TinyOS overhead makes it impossible to comply with the standard 
timing constraints.

*Issues that we think that have to be addressed on a future 
implementation of the ZB network layer in TinyOS 2.0*

- Synchronization (done by the lower layers)

- Timer granularities in heterogeneous mote networks

- Memory for allocating the routing tables and the routed frames

- Processing power for routing and still maintaining the protocol basic 
operations