Information and Communication Technology (ICT) is changing the way we live and has become an essential part of our life. With the advent of Internet of Things (IoT), and Wireless Sensor Networks (WSN) in particular, the number of devices that are networked is increasing exponentially over the years. With the increase in number of ICT devices ambient-energy harvesting has become a necessity. It is expected to liberate networked embedded systems from the use of batteries. Batteries provide us the necessary energy density and are able to power systems for a few months. Nowadays, rechargeable batteries can last a few years, typically 1 to 3 years, before they become unusable. On the other hand, energy harvesters such as photovoltaic panels, thermo generators, vibration harvesters etc., are permanent sources of energy, however with the handicap of having an unpredictable power output. The main challenge that is considered in this thesis is the perennial powering of embedded system with energy harvesting devices. This requires novel hardware and software mechanisms to deal with the unpredictability of such energy sources. As a first step towards our goal, the power output of energy harvesters has to be characterized under realistic conditions and thus in-situ measurements are performed for use throughout the thesis. Following this, efficient power-conversion and -management using clever circuit design techniques is investigated to make energy harvesters usable. Since the power output from the harvesters is unpredictable, a good energy prediction algorithm is required to make adjustments and circumvent low energy regimes. One has to look at replicable and generally applicable energy harvesting solutions to cater to as many environments and applications as possible. This means that energy harvesters have to be modelled, and a practical system has to be designed within a simulation environment to conduct detailed studies for understanding various situations and possibilities. This thesis has to also address the problem that in any given environment, there is a potential imbalance of the available energy across a network of nodes, since the amount of energy harvested by a node at a given location can be different compared to another node at a different location. Thus software and hardware algorithms need to be devised to ensure that the network remains unpartitioned. If data from a wireless sensor network is directed at an aggregation node or a gateway, power requirements for this node become crucial. Moreover, if data needs be transported over an Internet link, a system to collect data has to be devised and energy efficient mechanisms have to be implemented on such gateways. The gateway also requires mechanisms to send information to end-users in the most energy efficient manner. Data security under no circumstance can be compromised and at the same time, other performance objectives such as throughput and delay guarantees have to be satisfied. Thus mechanisms have to be devised to support security. Throughput and delay requirements may differ from one type of embedded system to another and depend on the availability of the system resources. Throughput maximization schemes are required to support a large variety of embedded platforms. Further, policies that enforce a throughput and delay optimum have to devised taking into account energy harvesting. Finally, real-time support with increased reliability, by exploiting spatial node-diversity, is required to address some of the important challenges in energy harvesting embedded systems wireless sensor networks. The results in this thesis address the gaps in the current state-of-art and proposes novel power management algorithms driven by empirical measurements. Let us now list a few major contributions of this thesis. (1) The unique circuit design for RF energy harvesting at -18dBm shows that the embedded system can transmit sensor data by ``accumulate and use’’ topologies. (2) In a multi-node environment, channel sensing is well known to avoid packet collisions. However, the impact of detection modes (energy and signal type) on energy consumption shows a novel way to manage harvested power in the framework of improved packet reception. (3) A perennially powered multi-node multi-hop energy harvesting testbed is setup with an advanced state-of-art energy harvesting protocol stack built over well-known MAC schemes such as BoxMAC-2 and routing algorithms such as AODV. (4) The testbed runs a multi input parameter based power management optimized distributed algorithm that boosts application performance. One such input parameter is the accurate energy prediction based on time series. (5) Suitability of well known cryptographic algorithms are studied for embedded platforms with available system resources. Key insights are gained about the best suitable scheme that does not compromise security at any cost. (6) Novel algorithms are constructed from existing ``throughput optimal’’ and ``delay optimal’’ policies for a flexible performance in WSN nodes. (7) Novel and sophisticated cooperative relaying schemes optimized for maximizing the packet reception and minimizing the delay are proposed to supersede simple relay selection schemes. (8) Delay and disruption of data due to link connectivity problems is well known. However, solution to data disruption due to energy availability advances the state-of-art and shows it is possible to reliability transfer substantially high amount of data for a given quantum of available energy. (9) A well-known modelling that adheres to energy conservation principle is applied innovatively to model energy harvesting sources and embedded systems. (10) The circuit model for a supercapacitor advances the state of art in battery-less storage of energy harvesting systems. This thesis is an initial step towards networked systems that do not require external energy sources, but instead draw their energy from their ambient environment. Such networks are referred to as Zero-Energy Networks.