One of the core features of IoT is global device discovery, allowing people to control and manage their connected devices anywhere across the internet. In pervasive IoT, device management becomes seamless and automated using AI for device pairing, content discovery, actions signaling, and command scheduling based on contextual information in a self-managed device paradigm. This chapter presents a novel handoff technique for smart speakers to ensure playback continuity when the users move around in a smart space consisting of multiple discoverable speakers. Current implementations include Spotify Connect, which is service specific, thus lacking in discovery robustness. The proposed handoff discovers at the device level instead of service level to improve device visibility and interoperability through a self-learning topology discovery (SLTD) method. For timely handoff, a rapid preemptive handoff method (RPF) is designed to optimize the TCP handshake mechanism and a content pre-caching technique is used to minimize playback gaps during the soft handoff.
TopIntroduction
Towards cloud and edge computing, devices discoverability is becoming more prevalent as ever for active device participation and network formation (Tianfield, 2018; Hong, 2017). In the early days, Digital Living Network Alliance (DLNA) is the go-to protocol for detecting the presence of devices but it has limited device support thus lacking in robustness. SNMPv3, a modern discovery protocol can identify devices up to sensors level, but the knowledge is local to the device's attributes without global awareness of surrounding networks (Yin, 2012). The inception of IoT further enhances the reachability of connected devices riding on popular application protocols like XMPP, CoAP, MQTT and some proprietary platforms like Apple Home, Google Home, and Amazon Alexa (Luzuriaga, 2015). Today, the vast connectivity among devices allows modern applications to take advantage of the device to device interoperability to enhance user experiences, like seamless connectivity for services continuity (Kang, 2013; N. Fathima, 2017; Barnaghi, 2016).
In the next-generation IoT, devices-to-devices communication is gradually replacing human-to-device communication towards A.I. driven IoT networks (Dawod, Georgakopoulos, Jayaraman, & Nirmalathas, 2019). User commands, which is traditionally invoked using IFTTT and Siri shortcuts (Vongchumyen, 2019) can now be delegated to A.I. for command prediction through probabilistic reasoning (Këpuska, 2018; Pham, 2018). One naïve example is the transition from user turning on a connected air purifier through MQTT publishing message is now automated by a PM2.5 sensor detecting bad air quality, then turning on the air purifier (Luzuriaga, 2015). The PM2.5 readings; called ambient data is used as a cue for reasoning; they can be extrapolated into the multi-dimensional aware system when cues like location, proximity, user identity, activity information are incorporated (Vishwakarma, 2019; Fortunato, 2019). Context-aware systems have taken great strides courtesy of modern IoT devices that are becoming more sensor-laden (Valarmathi, 2016). IoT devices can now benefit from ambient intelligence that unshackles them from limited locale constraints to seamlessly works and synchronise among IoT devices for more complex and coordinated tasks.
This chapter presents a novel technique for smart speakers’ handoff through user location tracking in indoor space. For the acoustic community, two of the most common challenges for desirable audio experience are; (1) audio fidelity and (2) area of coverage; which can now be addressed using location-aware speakers. Traditionally, music is streamed to a connected speaker talking to a nearby Access-Point (AP) in a smart space consisting of multiple independent speaker units. Currently, services like Spotify provides API for the discovery of all possible playback capable nodes; then, users can manually select the main speaker based on individual preferences. This chapter presents an automatic speaker selection technique to delegate the playback device selection to A.I. reasoning to fulfil the spatial requirement of balanced audio playback. The intuition here is music should be playback on the speaker that is closest to the listener in a smart space with multiple speakers connected in a near-mesh topology. Using spatial awareness, the music can be continuously played across a set of speakers by tracking user movements and changes in indoor positions. For example, among a speaker set Ss = {S1, S2, S3...Sn}; the algorithm chooses one active speaker (Sa) based on distance (d) between the user (Ux) and the speaker node (Si) at a discrete-time, ti. The handoff method ensures audio playback continuity towards an associated tracklet through speaker selection and hopping.