Mukesh Taneja

A framework for power saving in IoT networks

An IoT / M2M system may support large number of battery operated devices in addition to some mains operated devices. It is important to conserve energy of these battery operated constrained devices. An IoT / M2M Gateway used in this system is an intermediate node between IoT / M2M devices and an IoT / M2M Service Platform. It enables distributed analytics and helps to reduce traffic load in the network. This gateway could be stationary or mobile. In an IoT / M2M system, it becomes important to conserve energy of this Gateway as well. This paper proposes a framework to reduce power consumption of M2M / IoT devices as well as Gateway nodes. We buffer data at IoT Application, IoT Gateways and Devices to keep devices and Gateway nodes in sleep mode as long as possible. We allow computation of the duration to buffer this data using factors such as QoS requirements, predicted pattern of future IoT / M2M messages and congestion indicators from different network nodes. This potentially also allows intelligent aggregation of IoT messages at the Gateway node. We also enhance signaling mechanisms and present software building blocks for this framework. Mesh as well as Cellular access technologies are considered here.

A framework to support real-time applications over IEEE802.15.4 DSME

IEEE 802.15.4e has specified Deterministic and Synchronous Multichannel Extension (DSME) mode that helps to support real-time applications over wireless mesh networks. It supports a multi-superframe structure where each superframe during a beacon interval can potentially be assigned to a coordinating (or routing) device. Each superframe supports Contention Free Period (CFP) and Contention Access Period (CAP). IoT devices that send periodic data may want to use CFP while devices that send aperiodic data may want to use CAP (or need some dynamic way of using CFP for the period when such devices are sending data). In certain types of surveillance use cases, IoT devices may send low amount of data in normal scenarios (such as temperature reading, low resolution images or video) but these devices may start sending high volume of delay-sensitive data upon detection of certain types of events. For example, several video surveillance cameras may be activated upon detection of movement of people in an area. We need mechanisms that allow tradeoff between diverse needs of aperiodic and periodic IoT applications while at the same time, we need to be able to meet QoS requirements of delay-sensitive and other applications. A flexible resource management framework is proposed in this paper to achieve these goals. As packets move from an originating device to a destination (or coordinator) device in upward direction, various types of compensation parameters are computed and conveyed as part of MAC packets. The receiving (or coordinator) node uses these parameters to compute a compensation factor and conveys that to originating device in downward direction via protocols such as Constrained Application Protocol. This compensation factor is also carried in MAC packets along with other parameters in the upward direction. Intermediate nodes use compensation factor to do dynamic management of resources in DSME networks. It allows flexible resource management to meet requirements of diverse types of IoT applications in DSME networks.

An analytics framework to detect compromised IoT devices using mobility behavior

Certain security mechanisms assume that the end device is secured. In an IoT network, the IoT device itself could be compromised. An attacker could steal the device, gain access to it and use this for more damaging attacks. I propose an analytical framework where I specify certain mobility behavior indicators that are computed at network nodes and optionally at IoT devices. These are communicated to an analytics server using lightweight protocol enhancements specified here. IoT user specifies expected behavior using these indicators. Analytics server analyzes expected and observed values of these indicators and informs if it detects some unusual activity.

LTE-LPWA networks for IoT applications

Low Power Wide Area (LPWA) networks are characterized by support for long range operation and devices with low power and low throughput requirements. We consider a specific type of LPWA network (see LoRaWAN [2] for an example) where uplink packets sent from end devices are received by multiple intermediate gateways and forwarded to a network server. In the downlink direction, network server selects a gateway to send downlink packet(s) to a given device. We consider a scenario where uplink LPWA packets are aggregated at gateways using LTE networks. As multiple LPWA gateways receive copies of same uplink packet from an IoT device, these multiple copies (of same packet) are sent over the LTE network resulting in inefficient utilization of LTE network resources. It is not desirable especially when large number of IoT devices are deployed in an area or when some of these end devices start communicating lot more data on detection of some events. Also, if the network server selects a gateway for downlink traffic without having good information about intermediate LTE network elements, it could result in choosing a gateway that does not provide a good end-to-end path from network server to end device. We propose a traffic management framework that helps to reduce load in the LTE networks in uplink direction and help select a suitable LTE eNodeB and a gateway for downlink traffic in these LTE-LPWA networks. We allow support a software controller that allows IoT operators to specify different types of policies for uplink and downlink traffic management.

Policy based Automatic Neighbor Relation management for small cell networks

In a small cell network, it becomaes important to support Self-Organizing Network capabilities such as dynamic optimization. A User Equipment (UE) in a LTE network measures signal strength from neighboring access points (APs) and inform this to serving AP. Serving AP creates a Neighbor Cell List that has candidate APs for handover purpose. In this paper, we propose an enhanced Automatic Neighbor Relationship mechanism that considers load and QoS violation factors of the neighboring APs, in addition to received signal strength reports. We also present simulation results and show how it helps in doing policy based mobility management.

A framework for traffic management in IoT networks

Wireless networks for IoT applications support different types or classes of end devices. Each such class results in different uplink and downlink traffic behavior. It is important to identify suitable class for each end device. We first propose a generic framework for this purpose. We propose an element, called Software Controller, which learns device profile using variety of means such as information provided by the device itself, information provided by the associated IoT operator and contextual information using other sources. It can also use machine learning techniques to learn how a device might behave during certain period. Suitable resource management methods are to be associated with such classification schemes. We propose one such resource management method for 802.11ah type of networks. Next, we look at some traffic scenarios that may not be supported well by the existing device classes in some of these networks. Some IoT devices may always communicate low amount of data sporadically but some may need to communicate large amount of uplink or downlink (or bi-directional) data during certain time intervals. For example, an IoT device may need to measure (and report) certain parameters more frequently on detection of certain events, or a network server may want to set certain parameters or upgrade software at an IoT device during some time interval. It becomes important to control uplink / downlink communication opportunities and sleep interval at IoT devices in the network effectively. We propose a new device class and dynamic switching mechanism to handle such traffic scenarios effectively. We also include a software defined controller that provides for dynamic management of these communication opportunities at IoT devices and Access Points in the network.

A resource management framework for LTE-WLAN networks in high-speed trains

We consider a scenario where several WLAN users in a moving vehicle, such as in a train, are accessing internet using WLAN APs installed in the train. In addition to WLAN APs, there are also multiple mobile routers installed in that train. Here, a mobile router communicates with WLAN APs in the train and with LTE eNodeBs outside the train. A mobile router supports LTE User Equipment (UE) or modem type functionality to communicate with LTE eNodeB and Ethernet (or other) interfaces to communicate with WLAN APs. Packets from WLAN user devices get communicated over WLAN and LTE networks in this scenario. Even though a mobile router could support multiple LTE UEs in the same node, these UEs could offer different level of performance due to different channel and interference conditions in a high-speed train scenario. For uplink traffic from WLAN devices, it becomes important to select a suitable mobile router UE to communicate uplink data to LTE eNodeB and eventually to an application server in the internet. We first propose methods for uplink load balancing across different LTE UEs that are located in the same mobile router. We next consider uplink and downlink QoS management problem in such networks. LTE QoS scheduler runs at LTE eNodeB and considers variety of factors to allocate uplink and downlink resources to a UE. This scheduler treats each LTE UE as an individual UE and is not aware of multiple co-located UEs in a mobile router. We propose a network topology discovery method that allows LTE eNodeB to learn identity of multiple LTE UEs that are located on the same mobile router. We next use above methods to propose enhanced LTE QoS scheduling mechanisms to enable different types of resource management decisions in high-speed trains.